growth patterns and emplacement of the andesitic lava dome...

Geological Society, London, Memoirs

doi: 10.1144/GSL.MEM.2002.021.01.062002, v.21; p115-152.Geological Society, London, Memoirs

R. B. Watts, R. A. Herd, R. S. J. Sparks and S. R. Young Volcano, MontserratGrowth patterns and emplacement of the andesitic lava dome at Soufrière Hills

serviceEmail alerting to receive free e-mail alerts when new articles cite this article hereclick

requestPermission to seek permission to re-use all or part of this article hereclick

SubscribeCollection

to subscribe to Geological Society, London, Memoirs or the Lyellhereclick

Notes

© The Geological Society of London 2012

at University of Bristol Library on December 19, 2012http://mem.lyellcollection.org/Downloaded from

http://www.lyellcollection.org/cgi/alerts

http://www.geolsoc.org.uk/gsl/publications/page417.html

http://mem.lyellcollection.org//site/subscriptions/

http://mem.lyellcollection.org/

Growth patterns and emplacement of the andesitic lava dome at Soufri~re Hills Volcano, Montserrat

R. B. W A T T S l, R. A. H E R D 2, R. S. J. S P A RK S 1 & S. R. Y O U N G 2

1 Department o f Earth Sciences, Wills Memorial Building, University o f Bristol, Queens Road, Bristol BS8 1R J, UK

(e-mail: Rob. Watts@bris .ac.uk)

2 Montserrat Volcano Observatory, Mongo Hill, Montserrat, West Indies

Abstract: Eruption of the Soufri+re Hills Volcano on Montserrat allowed the detailed documentation of a Pel~an dome-forming eruption. Dome growth between November 1995 and March 1998 produced over 0.3km 3 of crystal-rich andesitic lava. Discharge rates gradually accelerated from < 1 m 3 s -1 during the first few months to >5 m 3 s -1 in the later stages. Early dome growth (November 1995 to September 1996) was dominated by the diffuse extrusion of large spines and mounds of blocky lava. A major dome collapse (17 September 1996) culminated in a magmatic explosive eruption, which unroofed the main conduit. Subsequent dome growth was dominated by the extrusion of broad lobes, here termed shear lobes. These lobes developed through a combination of exogenous and endogenous growth over many weeks, with movement accommodated along curved shear faults within the dome interior. Growth cycles were recognized, with each cycle initiated by the slow emplacement of a large shear lobe, constructing a steep flank on one sector of the dome. A growth spurt, heralded by the onset of intense hybrid seismicity, pushed the lobe rapidly out, triggering dome collapse. Extrusion of another lobe within the resulting collapse scar reconstructed the steep dome flanks prior to the next cycle.

In recent decades, phenomena observed during the growth of lava domes have been closely monitored, the most notable examples being at Mount St Helens, USA, between 1980 and 1986 (Swan- son et al. 1987), Mount Pinatubo, Philippines, in 1991-1992 (Daag et al. 1996) and Mount Unzen, Japan, in 1991-1995 (Nakada et al. 1999). As a result, many emplacement features and the processes controlling their formation have been described (e.g. Anderson & Fink 1992). The ongoing eruption of Soufri~re Hills Volcano on Montserrat has involved the construction of an andesitic lava dome, with alternating phases of growth and gravitational collapse (Young et al. 1998). The eruption required an intense scientific monitoring effort, due to the associated hazards and their threat to the local population, and this has yielded extensive data records on the eruption. These records have been used to distinguish patterns of growth and to develop a better understanding of the mechanisms controlling lava-dome eruptions (Sparks 1997; Voight et al. 1999; Melnik & Sparks 1999, 2002; Wylie et al. 1999; Sparks et al. 2000).

Because of near-daily helicopter observation flights during the eruption, an impressive photographic and video collection has accumulated, documenting the growth and collapse of the lava dome. Detailed monitoring of these morphological changes was also achieved using various surveying techniques, such that an accurate record of dome growth and magma production rate is available (Sparks et al. 1998). All of these data have been used, in conjunction with ground observations, theodolite, and electronic distance measurements, to produce maps detailing the complex development of the dome in time and space.

This paper describes the chronological evolution of the dome throughout the first episode of dome growth (November 1995 to March 1998) with the aid of maps and photographs, showing examples of the different structures that were extruded at particu- lar stages of the eruption. We discuss the relationships between the formation of these structures and the controlling mechanisms during magma ascent and emplacement of the dome. Under- standing of changing rates and styles of dome growth is vital in successful hazard assessment during dome-forming eruptions. We demonstrate that different growth styles are intimately associated with the generation of pyroclastic flows and the inception of explosive eruptions. We also develop an interpretation of the observations that attributes much of the morphological variation and behaviour to rheological stiffening of the magma caused by degassing and associated crystallization and also to deeper processes in the magma chamber, which periodically supplies pulses of fresh, gas-rich magma.

The terminology used is as follows. There was one lava dome extruded between 15 November 1995 and 10 March 1998. Extru- sion of a second dome began in November 1999, but is not dealt

with in any detail in this paper. Individual extrusions during the November 1995 to March 1998 period are termed lobes, and each is named by its date of first appearance.

Geological setting

Montserrat is a mountainous, diamond-shaped island, 16 km long and 9 km wide, that lies towards the northern end of the Lesser Antilles island arc. It is almost entirely volcanic in origin and is dominated by three volcanic centres. Recent 4~ dating (Harford et al. 2002) highlights a southerly shift in the focus of magmatism with time, from the low-lying Silver Hills (e. 2600 to 1200kaBp) in the north, through the Centre Hills (c. 950 to 550kaBP) and the South Soufri6re Hills-Soufri6re Hills complex (c. 160 ka to the present) in the south. The most recent activity has focused on the Soufri6re Hills region in the south-central sector of the island (Fig. 1). This activity has involved the sporadic growth and collapse of at least five andesitic lava domes and the formation of associated pyroclastic aprons surrounding these domes. The youngest and smallest dome, Castle Peak, was probably formed in a small eruption e. 350 years ago just prior to colonization of the island (Robertson et al. 2000; Harford et al. 2002). This dome nestled centrally within English's Crater (Fig. 2), a large structure breached to the east, which is believed to have formed by a c. 4 ka sector collapse (Roobol & Smith 1998). A moat thus circled Castle Peak dome on two-thirds of its circumference, with the final third facing into the eastward-trending Tar River valley. The 1995-1998 episode of dome growth involved the near-continuous extrusion of 300 x 106 m 3 of andesitic lava, constructing a complex Pel6an dome on top of Castle Peak dome and within English's Crater. A further 100 x 106 m 3 of lava has been extruded since November 1999 up to the time of writing (October 2000). A consequence of this eruptive activity has been the partial destruction and burial of both Castle Peak and the rim of English's Crater.

Eruption chronology

Since the onset ofphreatic activity on 18 July 1995, the eruption has progressed in an atypical manner in comparison with other well documented historical dome eruptions (Newhall & Melson 1983). From the initial phreatic period continuing through to the present day, a sequence of distinct eruptive phases has been experienced (see Table I). These phases highlight a fluctuating magma discharge

DRUITT, T. H. & KOKELAAR, B. P. (eds) 2002. The Eruption of Soufridre Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 115-152. 0435-4052/02/$15 �9 The Geological Society of London 2002. 115



116 R . B . WATTS E T AL.

Fig. 1. Pre-eruption map showing the location of andesitic domes (light shaded areas) of the Soufri+re Hills-South Soufri+re Hills complex in southern Montserrat. Dashed line marks the outline of the rim of English's Crater and the sides of the Tar River valley. Darker shaded areas are older uplifted pyroclastic sequences. Inset map shows the entire island.

Fig. 2. Early January 1996. View of Castle Peak (CP) sitting within English's Crater (EC), a c. 1 km diameter structure, looking west from above the Tar River valley. Pale lava on top of Castle Peak is new dome growth (D), and brown-stained vegetation results from early phreatic activity. Gages dome (G) is seen in background to right.

Table 1. Growth rates and associated surface/extrusive phenomena observed during the 8 eruptive stages of the current eruption

Stage Time period Growth rate (m 3 s - l) Surface phenomena and extrusive features

I

II

III

IV

V

VI

VII

VIII

18 July 1995 to 14 November 1995

15 November 1995 to 16 February 1996

16 February 1996 to 30 September 1996

1 October 1996 to 12 December 1996

13 December 1996 to 13 May 1997

14 May 1997 to 10 March 1998

11 March 1998 to mid-November 1999

Mid-November 1999 to date

Pre-dome

0.1-0.5

1-4 with daily spurts of >5 in July and August

0.5-2

2-4 with daily spurts of >5 in December and January

>5 with two phases of post-collapse explosive activity

No surface extrusion

2-5

Phreatic explosions

Spines, whaleback structures

Spines § megaspines, Type 1 + Type 2 shear lobes

Blocky lava § spines, endogenous activity

Megaspines + blocky lava, Type 1 + Type 2 shear lobes

Type 1 shear lobes, blocky lava

Sporadic dome collapse, sporadic explosions

Spines, Type 1 shear lobes + blocky lava



GROWTH PATTERNS AND DOME EMPLACEMENT 117

1100

1000,

900,

e- ._~ 800

"1-

(a)

30131 July collapse *,

~ Z $

to~ .~ �9 f

�9 ~'"" i 17 September *

collapse ~ ~

Cessation 25 June of growth collapse . t

. . ' J . .

26 December collapse

21 September / collapse --->-

,0% 11oo ~oo 300 ,ool ~o 60o ,oo-I~oo-~o-1ooo 1996 1997 1998

ones sending clouds of steam and dark ash aloft, which were blown westward towards the capital town of Plymouth. The first large phreatic explosion on 21 August initiated the first full evacuation of southern Montserrat. Of significance were the explosion craters of July, August and November, which formed a NNW-SSE alignment across Castle Peak (Fig. 4). Much of the subsequent magma discharge was focused along this zone, which suggests a fundamental fracture control on magma ascent. Extrusion of the first fresh lava occurred on the SW flanks of Castle Peak (Fig. 4), although it was initially interpreted as a cryptodome due to its highly oxidized nature. This small dome extruded on 25 September 1995 and continued to extrude for the next four or five days, generating small rockfalls prior to the growth stopping. A central spine of less oxidized material was extruded last, and degradation of the surrounding dome during October 1995 increased the prominence of this spine.

E % T- X

O a

120 (b)

100

80

60

40

20 II

o o,..i." lOO

V

[ IV ] .,.. "'~ III ..,... ,.-

. s

VI

I

VII

200 300 404 500 6()0 700 1800 960

1996 1997 1998 Eruption Days

1000

Fig. 3. (a) Graph showing the change in height of the active focus of growth throughout the eruption. Day one represents the onset of dome growth on 15 November 1995. (b) Graph showing the change in dome volume throughout the eruption (after Sparks et al. 1998 with more recent updates) and the different eruptive stages of the 1995-1998 period.

rate or pulses in magmatic activity that overprint a gradual escala- tion in eruptive vigour and background magma discharge rate (Sparks et al. 1998). As each phase of activity occurred, lava was extruded in a variety of styles and morphologies. Figure 3 details the volume of the dome and height of the active area of dome growth with time in the 1995-1998 episode of dome growth. Here we have divided the history of dome growth into stages, which are identified on the basis of prominent changes in dome growth patterns or significant volcanic and/or seismic events (Table 1). The main shifts in dome growth are given in Table 2 with notes on the different styles of growth during these shifts. Below we describe the dome evolution in chronological order, drawing attention to major styles of growth and morphological development. This paper does not consider in detail the new episode of dome growth, which started in November 1999, although similar patterns of dome growth are being observed to those described here.

Stage L" 18 July 1995 to 14 November 1995

The start of the eruption on the 18 July 1995 was heralded by a vigorous phreatic vent opening at the site of the poorly defined Langs Soufri+re on the NW flanks of Castle Peak (Fig. 4). Islanders were disturbed by a continuous roaring sound and the sight of a near-continuous jet of steam from within English's Crater. In later weeks, further new vents opened around Castle Peak (Fig. 4) and phreatic activity continued for over four months on a near-daily basis. Phreatic explosions were of a variable intensity, the larger

Stage H." 15 November 1995 to 16 February 1996

Phreatic activity continued throughout October and on until 15 November, when two small piles of fresh lava blocks were observed, one within the twin craters of the 18 July phreatic vent and the other between it and the September spine (Fig. 4). Growth of a new dome was confirmed, both by the presence of incandescent lava blocks and the onset of hybrid seismicity (Miller et al. 1998). The lava blocks were pale grey and generally <5 m in diameter, with larger, curving spines jutting out from the crater floor. The main growth occurred in the more southerly crater of July 1995, which was filled completely with new lava by the end of November. Spines were the main extrusive feature of this early stage of dome growth (Table 3), with typical widths of 30m and heights up to 35m. They exhibited a curved, outer surface covered by a thin breccia coating, with distinct subparallel grooves or striations running along their length. After several days of growth (typically at a few metres per day), spine collapse would occur due to gravitational stresses, forming piles of rock debris around the stubby base of the spine. Spines typically grew to an altitude of 810-860 m above sea level (a.s.1.) in this period of low magma discharge rates (0.1-0.5m 3 s-l). Growth of these spines was focused along the NNW-SSE zone defined by the early phreatic vents, particularly in the vicinity of the July 1995 craters. In Stage II, many spines emerged (Table 3) and grew on the western flanks of Castle Peak, such that the new dome developed as coales- cing piles of spine debris (Fig. 5a). Fumarolic activity commonly occurred at the edges of these piles with associated rockfall activity increasing as the dome grew. A prominent feature of the freshly extruded lava was the apparent lack of flow structures and near- solid appearance of the blocks and spines.

An interesting feature, formed between 24 January and 6 Febru- ary, was the formation of whaleback structures coincident with a period of seismic tremor and slightly raised extrusion rates of around 1 m 3 s - l . These elongate bodies extruded individually as extremely viscous lava from the NNW-SSE vent pattern in different directions (Fig. 5b). Each whaleback was pushed across the surface of Castle Peak by continued magmatic pressure (in a manner analogous to toothpaste being squeezed out of a tube). Whale- back structures would reach up to 200m long, 30m wide and 35 m high, and exhibit a smooth surface with a grooved appearance and striations aligned parallel with the direction of extrusion (Fig. 6a, b). Growth of each whaleback structure occurred sporadically at an estimated 20-30 m day -1 and continued to move for up to a week prior to growth stagnating and the subsequent break-up of the structure into a chaotic jumble of blocks (Fig. 7). During growth, as the whaleback pushed forward, rockfalls spalled off its steep headwall, exposing incandescent lava within the interior. Whalebacks are previously undocumented, although they have been witnessed during other eruptions (e.g. Santiaguito dome in Guatemala, W. I. Rose, pers. comm.). Less distinct rubbly lava also extruded in January 1996, although poor visibility prevented detailed documentation.



118 R . B . WATTS ET AL.

Table 2. Chronology of activity in the 1995-1998 period of dome growth at the SoufriOre Hills Volcano

Date Activity

25-29 Sep. 1995

15Nov. 1995

24 Jan. 1996 to 3 Feb. 1996

15-18 Feb. 1996

Mid-Feb. to early April 1996

Late April 1996

1-3 June 1996

22-25 June 1996

12-15 July 1996

Mid-July to early Sep. 1996

17 Sep. 1996

1 Oct. 1996

Early Nov. to early Dec. 1996

13Dec. to 24Dec. 1996

25Dec. 1996 to 10 Jan. 1997

Mid Jan. to late Jan. 1997

Early Feb. to 26 March 1997

27 March to mid-May 1997

16May 1997

17May to early July 1997

Mid-July to late July 1997

Late July to 12 Aug. 1997

13Aug. 1997

8 Sep. 1997

21Sep. 1997

22 Oct. to 2 Nov. 1997

3-6 Nov. 1997

Mid-Nov. to 25 Dec. 1997

26 Dec. 1997

27 Dec. 1997 to late Feb. 1998

1-10 March 1998

First juvenile lava extruded as small pile of blocks and spines on SW flanks of Castle Peak, but extrusion rapidly stagnated.

Onset of continuous extrusion of juvenile lava as spines on NW flanks of Castle Peak.

Extrusion of northern whaleback over several days followed by southern whaleback.

Extrusion of southeastern whaleback.

Repeated extrusion of large vertical spines in central area of Castle Peak, spine collapse produces first pyroclastic flows down Tar River valley.

Growth spurt preceded by hybrid earthquake swarm, forming shear lobe of blocky lava (i.e. 25 April 1996 lobe) at summit directed to NE flanks.

Extrusion of megaspine to form northern peak.

Extrusion of megaspine to form southern peak.

Extrusion of megaspine to form southwestern peak.

Vigorous extrusion of shear lobes of blocky lava towards NE dome flanks during growth spurts triggering dome collapses on 29-31 July, 11-14 Aug. and 2-3 Sep.

Directed explosion following major dome collapse down Tar River valley removing one-third of dome volume.

Onset of renewed dome growth (i.e. 1 Oct. 1996 lobe) at base of explosion crater at low extrusion rate.

Intense seismicity accompanying minor dome growth, suggesting period of endogenous activity. Localized doming on southern dome flanks.

Rapid extrusion of 13 Dec. 1996 megaspine along shear zone in southeast sector. This megaspine is quickly overwhelmed by blocky lava generating dome collapse on 19 Dec.

Rapid extrusion of 25 Dec. 1996 lobe forming 'pancake' morphology in central part of dome following localized endogenous doming. This rapidly overwhelms 13 Dec. 1996 lobe and 1 Oct. 1996 lobe.

Rapid blocky growth from central area directed across SE flanks, generating semi-continuous pyroclastic flow activity and large collapses on 16, 19 and 20 Jan.

Blocky lava of 21 Jan. 1997 lobe directed down southeastern and eastern flanks from central area forming a conical dome morphology.

Growth of 27 March 1997 lobe guided southwards triggering collapse of southern flanks on 30-31 March and 11 April leading to inundation of White River valley.

Large vertical spine extruded at dome summit.

Growth of 17 May 1997 lobe across northern flanks triggering dome collapse down northern flanks on 25 June followed by regrowth of 27 June 1997 lobe to north.

Stagnation of active lobe, growth slightly switched to NW flanks (14 July 1997 lobe).

Repeated small collapses down western flanks lead to major collapse on 3 Aug. Entire 14 July 1997 lobe collapsed triggering period of 13 repetitive Vulcanian explosions.

Renewed blocky growth in crater followed by formation of active lobe to the west.

Stagnation of 13 Aug. 1997 lobe with fresh lobe growth directed northwards. Rockfall and pyroclastic flow activity completely fills Mosquito Ghaut.

Rapid extrusion of 8 Sep. 1997 lobe to the north triggers major dome collapse directed down Tuitt's Ghaut forming amphitheatre-shaped edifice. This triggers a period of 75 Vulcanian explosions (22 Sep. to 21 Oct.).

Renewed blocky growth in crater developing into northward growing 22 Oct. 1997 lobe.

Stagnation of 22 Oct. 1997 lobe contemporaneous with three collapses concentrated on the southern flanks attributed to switch in active lobe to the south.

Continuous growth of southerly directed 4 Nov. 1997 lobe constructing a large peak straddling former Galway's Wall area.

Major debris avalanche and sector collapse removing the entire southern flanks (i.e. 4 Nov. 1997 lobe) formed in previous two months. Violent pyroclastic density currents.

Redevelopment of southerly directed lobe (i.e. 27 Dec. 1997 lobe) rebuilding the large peak on the southern flanks.

Cessation of first period of dome growth signalled by 50m high spine at summit of 27 Dec. 1997 lobe.



Fig. 4. Map of English's Crater on 25 November 1995 showing the location of phreatic vents (with dates when they first appeared) and the initial extrusions of fresh andesitic lava in the early eruptive stages. Note that the 21 August phreatic vent opened up along a fracture-controlled line of smaller vents. Farrell's Wall, Gages Wall and Galway's Wall represent different sectors of the steep avalanche scar defined by English's Crater. The rim of the crater is marked by a dashed line. Peaks A B and C on this rim were used as prominent topographic features in surveys of the dome. Map co-ordinates are part of the Montserrat Grid System.

380500

Chances ~ v 25 Sept Dorr Peak e" ~ . ~ , , . ~ ~ - - ~ ~ ' ~ ' - ~ ~ e + tal u s

0 100m I I

23 N O V E M B E R 1995

Aug

381400

LEGEND

~ Explosion pit

O Avalanche scar i O Sept 1995

dome

Nov 1995 dome

Contour Interval - 100 feet 1 ft=0.3048 m \

Dome Volume = 0.1x106 m 3 3 -1 Extrus ion R a t e = ~ 0 . 1 7 m s

Table 3. Theodolite data collected during Stages H and I l l of the eruption to monitor the height and growth rate o f spines extruded at this time

1847600

1846600

Date (eruption day) Theodolite location Height of spine Height of summit Spine growth rate (m a.s.l.) (m a.s.1.) (m day -1)

15/11/95 (1) Long Ground 5/12/95 (20) Long Ground 814 10/12/95 (25) Long Ground 805 17/12/95 (32) Whites 805 20/12/95 (35) Whites 825

24/12/95 (39) Whites 817 26/12/95 (41) Whites 814 28/12/95 (43) Whites 812 30/12/95 (45) Whites 817 3/1/96 (49) Whites 810 8/1/96 (54) Whites 814 11/1/96 (57) Long Ground 842 22/1/96 (68) Photo method* 5/2/96 (82) Photo method* 21/2/96 (98) Whites 11/3/96 (117) Observatory 19/3/96 (125) Whites 22/3/96 (128) Whites 857 25/3/96 (131) Whites 874 26/3/96 (132) Whites 885

29/3/96 (135) Whites 30/3/96 (136) Whites 862 5/4/96(142) Long Ground 883

18/4/96 (155) Long Ground 905 20/4/96 (157) Long Ground 30/4/96 (167) Whites 929

760

823 825 821 866 847

855

903

>7 Cathedral spine

9 Highest of 3 spines

2.5

4 >9

9 5

11 Spine width = 35 m

14 Spine width = 44 m

* Measurements obtained by a technique used to calculate changes in dome volume through comparison of photographs taken from the same location. Dates in European format (5/12/95--5 December 1995). a.s.l., aboe sea level.



120 R. B. WATTS E T AL.

Fig. 5. Sequence of maps showing gradual development of the dome during the initial three months of dome growth leading to onset of pyroclastic flows at the end of March 1996. Shaded area marks the new dome growth. X-Y represents the line of section used in Figure 7.




Fig. 6. (a) 2 February 1996. View looking west above Castle Peak showing two whaleback structures on the new dome: the southern whaleback is marked S and the northern whaleback N (see Fig. 5b). Note the steep walls of English's Crater (EC) and the dome of Chances Peak (ChP) and telecommunications tower in the centre background. Gages Wall (G) is on the right with Plymouth Town (P) in the distance. (b) 1 February 1996. View looking east from crater rim near Chances Peak. Elongate structure in centre is the southern whaleback (S), which extruded subhorizontally away from the central area over several days (see Fig. 5b). Note the smooth outer surface of the whaleback, with rocks spilling off the leading edge on the right. Loose blocks sitting on its top are carried along during growth. CP is the main spine of Castle Peak dome.

Stage IlL" 17 February 1996 to 30 September 1996

In mid-February, a marked rise in the average background extrusion rate (r 2 m 3 s -1) was experienced at the start of this period, following a volcanotectonic earthquake swarm on l l February 1996. There was also a steady increase in dome height, which reached 960m a.s.1, by the end of June 1996 (Fig. 3a). Spectacular spines with prominent vertical striations scored along their smooth, outer surfaces, and exhibiting characteristic curved-horn shapes

were commonplace throughout March and April (Fig. 8a, b). Growth rates of these spines averaged 6 - 7 m day -1, sometimes over 10m day -1, attaining heights of c. 40m before collapsing (Table 3). Spine growth was generally concentrated in the central area of Castle Peak, a slight southerly shift from the initial growth area of mid-November 1995. The continuous formation of spine debris enlarged the area covered by the new dome. By the end of March, the entire western half of Castle Peak had been overwhelmed by fresh lava (Fig. 5c) and blocky talus was rapidly



122 R.B. WATTS ET AL.

Fig. 7. Schematic diagram showing the development of whaleback structures in January 1996, along section X-Y in Figure 5b. (a) 15 January 1996. Dome growth occurring as spine extrusion and gravitational collapse. (b) 24 January 1996. Whaleback structure extrudes northwards away from central area averaging 30 m day -l . (e) 29 January 1996. Growth of northern whaleback stops and another whaleback starts to drag across the dome surface southwards away from the centre at 20-30m day 1. (d) 5 February 1996. Both whalebacks have stagnated and gradually break up. Spine growth and collapse recommences around summit area.

filling the moat of English's Crater. Rockfalls were no longer confined by the walls of English's Crater on the NE flanks, and they developed into small pyroclastic flows down the upper parts of the Tar River valley.

The pulsatory character and most common style of dome growth can be illustrated by. observations in the second half of April. Two spines were extruded around 14-16 April from the July 1995 crater area coincident with a period of elevated hybrid- earthquake seismicity. The spines toppled over, one to the west and the other to the east. While the westerly spine remained stagnant, the easterly spine started to break up on 25 April as new lava started to extrude. This blocky lobe then expanded to the NE and formed a steep headwall at its advancing front from which rockfalls cascaded down into the Tar River valley. The flow front position

stagnated when it reached the steep edge of Castle Peak with generation of rockfalls at the flow front matching supply of lava from the lobe. The flow front developed a furrowed appearance of ridges and chutes related to repeated generation of erosive rockfalls spilling down from the summit of the lobe. This blocky lobe, here termed a shear lobe to highlight this process of directed extrusion (Fig. 9a), continued to produce lava until 2 May, when growth became more focused around the summit. This formation of shear lobes was repeated many times throughout the eruption although the style of lobe development was markedly different in the later eruptive stages (Fig. 9b, c), allowing a classification between early- stage Type 2 lobes and late-stage Type 1 lobes. A feature common to both types was that the summit of a lobe was commonly shifted away from the vent area during growth, giving an illusion of shifting vent positions as an individual lobe developed. A similar illusion occurred when a lobe stagnated and a fresh lobe was initiated, growing in a different direction from the previous lobe. Formation of Type 2 lobes was apparent after extended periods of stagnation, while Type 1 lobes predominated during periods of more steady growth, although fluctuations in discharge rate were evident during the development of both structures.

From June to July 1996, noticeable shifts in the focus of dome growth followed the slow extrusion of broad features (up to 100m wide) that are best described as fault-bounded megaspines (Fig. 10a, b). A megaspine is characterized by two contrasting parts. One side of a megaspine consists of a smooth, striated and curving wall which is interpreted as moving along a large fault in the dome. The other side consists of a headwall of massive, blocky material that breaks up during growth. A megaspine grows by upward or subhorizontal movement along the fault structure with lava blocks spalling off the main headwall as growth occurs (Fig. 10c). Em- placement of such a large structure often stopped after a few days, with renewed activity taking place in another localized part of the dome. The most notable examples of megaspine growth occurred during 1-3 June (to the north), 22-25 June (to the south) and 12-15 July 1996 (to the SW). Each of these extrusions formed a prominent peak on the dome and originated from the same central focus of growth. At this central focus, extrusion of fresh lava was guided along a curvilinear shear fault that directed the lava in a specific direction, sometimes over a hundred metres away from the previous growth area (Fig. 11).

By mid-July 1996, the new dome was a substantial size (c. 30 x 106m 3) and had multiple peaks due to the repeated formation of megaspines. Vigorous spurts of dome growth were also evident at this time in pulses lasting several hours, at estimated discharge rates of >5 m 3 s -1. The focus of growth was located in the central area, with the 17 July 1996 lobe directing fresh lava down the NE flanks of Castle Peak and the upper reaches of Tar River valley. At these rapid growth rates, distinctive piles of blocky lava (c. 4-5 m diameter) were extruded, exhibiting a curvilinear shape with occasional larger spines projecting out. The coincidence of this vigorous growth and the focus of growth directing lava to the NE resulted in a series of major dome collapses producing large pyroclastic flows on 29-31 July, I I August, 20-21 August and 2-3 September 1996. This involved the repeated collapse and reconstruction of the NE dome flanks (Fig. 12a, b) and the gradual inundation of the entire Tar River valley (Cole et al. 2002). These repeated collapses led to a decrease in the height of the active area of growth (Fig. 3a). Renewed growth was always focused at the base of the spoon-shaped scar formed by each collapse, often accompanied by a vigorous, semi-continuous ash plume from the central area. A new lava lobe, often bounded by a shear surface at the backwall, would then extrude as large, curved blocks that rapidly filled and engulfed the entire scoop (Fig. 12c). The lava lobe always expanded asymmetrically towards the open side of a previous collapse scar, guiding it away from the vent area. A particularly long episode of near-continuous collapse (c. 9 hours) occurred on 17 September, culminating in sub-Plinian magmatic explosive activity (Robertson et al. 1998). This sustained collapse removed over one-third of the new dome (c. 11.8 x 106 m 3 volume non-dense rock equivalent, all volumes quoted here are collapse volumes calculated from deposit volumes and/or collapse scar




Fig. 8. (a) 9 April 1996. View looking north across English's Crater and the new dome, with the northeastern flanks and airport in the background. Large spine of fresh lava (P) is c. 40 m high and c. 35 m broad. Broad spine in centre (Q) is the main spine of Castle Peak dome. Note Galway's Wall (GW) in foreground. (b) 12 April 1996. Typical curved-horn spine extruded in Stages II and III. Note the semi-cylindrical shape with smooth and striated outer surface. Other half of spine is massive lava of the spine interior that spalls off to form a basal skirt of debris around the base of the spine. This spine is estimated to be 40 m high and 35 m broad. Note Perches Mountain (PM) in background to the right.

volumes (Calder et al. 2002), and a substantial portion of the underlying Castle Peak dome, leaving a large steep-sided scar in the central area open to the east (Figs 13a and 14a).

Stage IV." 1 October 1996 to 12 December 1996

Extrusion of fresh lava did not resume for two weeks following the 17 September 1996 explosive eruption, which was estimated to have

reamed out the conduit to a depth of 4 km (Robertson et al. 1998). This event involved substantial widening of the upper conduit, with abundant ballistic ejecta of vent-wall breccia, hydrothermally altered rocks of Castle Peak and dense, juvenile blocks thought to have originated in the upper conduit (Robertson et al. 1998). The explosion was directed to the east as evidenced by strong asym- metrical distribution of ballistic ejecta, and this activity is believed to have flared open the upper conduit. Lava started to extrude at the base of the scar on 1 October with an initial discharge rate of 1.8 m 3 s -1. The focus of upwelling was noticeably shifted c. 150 m to



124 R.B. WATTS ET AL.

Fig. 9. Characteristic features of Type 1 and Type 2 shear lobes shown in schematic form. (a) Development of a Type 2 shear lobe: following an extended period of stagnation, a fresh pulse of magma pushed out a viscous plug that was emplaced as a large spine or megaspine. The hotter magma that forced out the plug continued to ascend along another shear fault that detached from the conduit wall and magma extruded in the direction of least resistance (see Fig. 11). During extrusion, the lobe broke up into curving blocks with the highest point often displaced laterally away from the conduit, giving the illusion of shifting vent locations when a new lobe moved in another direction. The steep flow front advanced until it reached a steep slope (e.g. the margin of Castle Peak or the rim of English's Crater) where the front stagnated and generated rockfalls and pyroclastic flows as lava was supplied from behind. (b) Early stage development of a Type 1 shear lobe: following shorter periods of stagnation, a new shear lobe extruded with a large, coherent, smooth and striated upper surface with a broad headwall of massive lava. The upper surface developed a broad, semi-cylindrical shape supported by the surrounding dome flanks. This switch in the focus of activity and early stage development of a new lobe often triggered a dome collapse, e.g. 4 November 1997. (e) Late stage development of a Type 1 shear lobe: following stagnation of the viscous plug, hotter and more ductile lava would rise up along the same shear zone. Broad, curving spines extruded and broke up at the rear of the lobe and this activity alternated with the injection of magma into the core of the lobe. This latter process expanded the summit area and triggered rockfalls off the leading edge of the lobe. The repeated nature of these processes over several weeks developed a conical summit with a broad skirt of blocky talus. Formation of a Type 1 lobe was most apparent during periods of steady-state growth when only minor fluctuations in the average discharge rate were experienced.

the east in comparison to the location of spines and upwellings before 17 September 1996. From this time, the dimensions of lobes generally became substantially greater. This temporary shift in growth foci and the larger dimensions of subsequent lobes are attributed to the widening of the upper conduit asymmetrically to the east.

Initially, the new lobe consisted of a slab of smooth lava over- lying loose talus (Fig. 14a, b). The morphology of this lava exhibited a transition over several days from smooth (Fig. 14b) to an unusual darker, rubbly surface (Fig. 15a) and eventually to the blocky and spiny appearance that had previously characterized the dome. This period of activity shows an excellent example of the morphology formed by new growth infilling the scar after a major collapse. The early growth of the 1 October 1996 lobe was also the first example in this eruption of a lava morphology affected by lateral spreading (Fig. 15a). As growth continued, the lobe gradually filled the scar at a decreasing discharge rate and had apparently stagnated by 20 October 1996. Renewed extrusion, still at a reduced rate, took place on 22 October, but focused only on the central part of the lobe. This formed a central raised area of blocks and small spines surrounded by the lower abandoned sectors of the lobe (Fig. 13b).

In early November 1996, intense shallow earthquake swarms occurred, dome growth rates dwindled to <1 m3s -1, and dome

height stagnated at 900 m a.s.1. By mid-November, further earthquake swarms triggered landslides off the steep outer face of Galway's Wail on the southern rim of English's Crater (Fig. 15b). This apparent intrusion into the dome, causing localized uplift of over 30 m on the southern dome flanks, was the first clear evidence for endogenous activity during the eruption. This raised levels of concern that the threat of sudden collapse of Galway's Wall might trigger rapid decompression of the dome interior (Young et al. 2002).

Stage V: 13 December 1996 to 13 M a y 1997

The crisis related to the instability of Galway's Wall was temporarily relieved on 13 December 1996 when the southeastern margin of the 1 October 1996 lobe was uplifted along a major ductile shear zone bounded by a steep, striated fault with a trend of 110 ~ The 13 December 1996 lobe (also known as the 'Venus' lobe) initially extruded as a megaspine near the zone of endogenous uplift. This fault-bounded feature extruded rapidly in a southeasterly direction overnight and continued growth along the same trend for several days, forming a single entity at least 150 m long, 100 m wide and 100 m high (Figs 16a and 17a). Within days, the megaspine had been




Fig. 10. (a) and (b) Maps highlighting the switches in the focus of activity during June and July 1996 following the emplacement of megaspines. Dotted lines represent approximate boundary between massive lava and loose talus blocks (for clarity, this distinction is not marked in later maps). (c) 19 August 1996, looking east from summit of Chances Peak. Megaspine extruded in early July 1996 (c. 40 m high and c. 100 m broad) showing a smooth, curved northern face to the left and massive, crystalline lava breaking off its broad headwall (H) as large blocks to the right. This large feature was subvertically emplaced in the northwest sector of English's Crater (see a) and remained a prominent peak on the dome for the following six months. Also in view is another megaspine (S) emplaced on the southern sector of the dome in late June 1996 (see b).

overwhelmed by the pulse of magma that had initiated its extrusion. Near-continuous rapid extrusion of blocky lava (c. 4-6 m 3 s -1) was guided along the same southeasterly directed shear fault that extruded the megaspine, and pyroclastic flows spilled off the SE dome flanks.

On 25 December 1996 a pronounced pulse of activity, heralded by localized endogenous doming and episodes of banded seismic tremor (Miller et al. 1998), led to the emplacement of the 25 December 1996 lobe, also known as the 'Santa' lobe (Fig. 16b). This new lobe initially punched through the central summit of the




• u Fig. 11. Schematic diagram illustrating the emplacement of a megaspine and the switching of activity that was apparent in June and July 1996. Diagrams represent a cross-section along the line X-Y in Figure 10b. (a) 30 June 1996. This was a period of very slow growth with only minor rockfall activity occurring at the summit. Lava within the upper conduit is crystallizing and forming a near-solid plug within the dome. (b) 13 July 1996. Increased seismic activity heralds a pulse of fresh magma through the conduit. This increased pressurization results in the lava plug being pushed out along a curved shear fault within the dome. This plug is emplaced as a megaspine at the summit producing rockfalls that break away from its headwall during growth. (c) 27 July 1996. Following emplacement of the megaspine, the hotter magma beneath is redirected along another shear fault that provides an easier pathway to the surface. The rapid rise from depth of hot magma at the head of the fresh magma pulse does not allow enough time for a plug to develop in the upper conduit. The 17 July 1996 lobe was a typical example of a Type 2 shear lobe comprising large blocks and stubby spines sourcing pyroclastic flows.

13 December 1996 lobe at a vigorous pace, with field observations suggesting a discharge rate between 6 and 9 m 3 s -1. This fresh lobe spread out laterally across the summit to form pancake-shaped lava on top of the 13 December lobe. Notably, some of the most spectacular examples of incandescence during the night-time and even daytime were experienced during this period associated with rapid discharge rates (Fig. 17b). In the following two weeks, this lobe rapidly overwhelmed both the 1 October 1996 lobe and the 13 December 1996 lobe through lateral growth (Fig. 18). The 25 December 1996 lobe provides the best example of the more fluid- like behaviour exhibited by the lava when dome growth was more vigorous. The top third of the dome (c. 50-70 m) on 28 December 1996 consisted of a front of incandescent blocky lava which con-

tinually generated incandescent rockfalls down the lower two-thirds of the dome. The circular plan-form and slightly raised central summit of this lobe indicated that lava was extruded in the summit region but was able to spread laterally, forming an overall pancake- like morphology. Emplacement of this somewhat more fluid lava contrasts with the predominant extrusion of spines and lobes along ductile shear faults.

The 25 December 1996 lobe initially spread symmetrically, spilling lava down the entire eastern flanks, but by mid-January its advance had become more focused towards the SE. The continued advance of the lobe guided lava down the SE dome flanks and triggered large pyroclastic flows down the SE side of the Tar River valley on 9 16 and 19-20 January 1997. With the northern half of the 25 December 1996 lobe now effectively abandoned, subsequent blocky growth then rapidly infilled the SE-facing scar, producing a series of small collapses (Figs 19c and 20a, b).

By early February 1997, Castle Peak had been partially destroyed and completely buried by rockfalls and pyroclastic flows, although the discharge rate had declined to more moderate levels (2-4m 3 s-l). A slight switch in steady growth of the southeasterly directed 21 January 1997 lobe to a more easterly trend, focused rockfall activity down the eastern flanks throughout late February and early March. The conical dome that developed in the latter half of March illustrated well the effect of rockfalls on dome morphology. This period provided a very good example of the evolution of an asymmetric Type 1 shear lobe over an extended period of sustained growth. Large, curving spines of fresh lava would extrude away from the rear of the lobe, then gradually push forward and break up, spilling blocks onto the lobe summit. This activity alternated with magma pushing into the molten core of the lobe, expanding this area and triggering rockfalls off the lobe headwall (Fig. 9c). A swath of rockfalls from both collapse of spine debris and disintegration of the lobe headwall gradually formed a lobe with a furrowed appearance as rockfall chutes developed. This repeated process of alternating spine growth and magma injection into the core gradually formed a lobe with a broad, conical form (Figs 21a and 22a).

The end of March witnessed the extrusion of a remarkable structure, formed in the early development of the 27 March 1997 lobe (also known as the 'Easter' lobe). The growth of this lobe occurred at a time of good visibility and its development was well documented. The appearance of this structure immediately preceded a dome collapse on 30-31 March (c. 3.6 x 106 m 3 non-DRE deposit volume) focused on the southern flanks; this destabilization is attributed to the initial growth of the 27 March 1997 lobe. A broad mass of lava started to project out in a subvertical manner on 27 March, originating from about 50-70 m below the summit area of the conical dome lobe extruded throughout March. Its movement took place in a stick-slip fashion at estimated rates of 25-30m day -1, with movement of the lobe accommodated by a southerly directed ductile shear fault in the dome. By 3 April, the 27 March 1997 lobe exhibited a smooth, yet striated, curved upper surface (c. 100 m long and c. 120 m wide) and a near-vertical headwall of massive lava almost c. 150 m high (Figs 21b and 22b).

Growth of this structure involved gradual rotation guided along by the shear fault surface forming a semi-cylindrical cross-section. Vigorous degassing emanated from around the horseshoe-shaped boundary zone between the lobe and the inactive parts of the dome. On initial extrusion, the smooth backwall was directed subvertically, and as growth continued this changed down into a more subhorizontal position (Fig. 22c). This mode of extrusion can be closely correlated to the shear lobes of blocky lava evident in mid- 1996. In the case of the 27 March 1997 lobe, however, the smooth, semi-cylindrical upper surface remained as one large coherent structure during extrusion instead of breaking up into smaller, curving blocks that gradually rotated forwards during growth (Fig. 9b). This difference may be partly attributed to the fact that the 27 March 1997 lobe was buttressed and supported by much larger surrounding dome flanks than the blocky lobes apparent in the earlier eruptive stages.

The 27 March 1997 lobe continued to grow in the same manner, bulldozing through the southern flanks of the new dome and




Fig. 12. (a) and (b) Maps highlighting the scales and locations of collapse scars of 12 August 1996 and 2 September 1996, both formed as a result of the growth spurts prevalent in the Stage III period. Shaded area represents the new dome. Legend as in Figure 5. (c) 25 August 1996. View from point Y on (a) looking at NE flanks of new dome. Fresh lava blocks are extruding from the summit area (marked by arrow) and being directed down the NE flanks. Such growth was typical during spurts of activity that were commonly experienced during late July and August 1996. Note the ash-venting near the focus of extrusion at the rear of the lobe. Dashed line marks the scar rim from the 11 August 1996 collapse within which fresh lava of the 12 August 1996 lobe has filled up the scar in a two-week period. CP marks the prominent two-pronged spine of Castle Peak.

destroying the upper rampar ts of Galway 's Wall (Fig. 23a). Large blocks of massive lava spalled off the headwal l of this lobe as it je rked forward to generate rockfalls and pyroclastic flows down Galway 's Wall and into the Whi te River valley. With in a week, dome talus and pyroclast ic flows had overwhelmed wha t remained

of Galway 's Wall. Fol lowing another dome collapse (c. 3 x 106 m 3 deposit volume) on the southern flanks on 11 April , a well developed lobe was observed extruding in a similar m a n n e r and t rend as the 27 M a r c h 1997 lobe. The smooth backwall of this 13 April 1997 lobe, however , was steeply angled at c. 45 ~ whilst the leading edge




Fig. 13. (a) and (b) Maps highlighting the scale and configuration of the 1 October 1996 lobe growth following the 17 September 1996 explosion. Light-shaded areas mark lava extruded prior to the September explosion; darker shaded areas mark the 1 October 1996 lobe. Legend as in Figure 5.

continued to push outward and break up during growth. Slow growth of this lobe continued for a few weeks until stagnation in mid-May, when the headwall of massive lava remained as the steep, upper ramparts on the southern dome flanks. A broad skirt of talus fanned away from these ramparts (Fig. 23b) onto the deposits from the associated pyroclastic activity, which had destroyed most of the White River valley (Cole et al. 2002).

S t a g e VI." 14 M a y 1997 to 10 M a r c h 1998

This period commenced with a gradually accelerating background rate of extrusion (c. 4 -5m 3 s -1) and an increase in dome height to 1000m a.s.1. Between 14 and 17 May 1997, a distinct switch in the focus of growth took place, returning to the central area with an impressive c. 50-m-high vertical spine evident at the summit on 16 May (Fig. 24a, b). Within two days, this spine had collapsed and lava extrusion was guided in a northerly direction, with rockfalls spilling down a broad swath covering the southeastern-to-eastern sector and on around to the NW flanks (Fig. 24c). Despite limited visibility, the stark contrast between the shutdown of activity on the southern flanks and the broad spread of activity across the entire northern flanks indicated the development of the 17 May 1997 lobe directing lava to the north. This rapid switch in dome growth also indicated that vertical extrusion of lava was only evident between major switches in activity, with the preferred mode of emplacement taking place through subhorizontal shear lobes. This semi- continuous rockfall activity continued unabated at a steady rate for several weeks, broadening the dome flanks against the northern walls of English's Crater. By early June 1997, talus was level with Farrell's Wall on the northern rim, and rockfalls tumbled directly down the northern dome flanks into Tuitt's Ghaut and Mosquito Ghaut. The ensuing weeks involved increasing pyroclastic flow activity down these ghauts, threatening the communities of the northeastern slopes of the volcano (Loughlin et al. 2002).

On 22 June 1997, a marked change in seismicity occurred, with an 8-hour cyclic pattern of intense hybrid earthquake swarms and associated cycles of ground deformation recorded by near-vent tiltmeters (Voight et al. 1999). This period of increased pressurization continued for several days and culminated in the rapid destabilization of the entire 17 May 1997 lobe. A major dome collapse (c. 6.4 • 106m 3 deposit volume) on 25 June resulted in the first fatalities of the crisis (Loughlin et al. 2002). Large pyroclastic flows were funnelled down Mosquito Ghaut, reaching about 6 km to the NE, lowering the dome summit by 100m and excavating a steep- sided scar on the northern dome flanks. Within a few days the scar was refilling with blocky lava and stubby spines to construct the 27 June 1997 lobe. By 10 July, a broad headwall of massive lava had risen up within the scar indicating the northerly directed lobe had re-established itself (Fig. 25a).

A gradual switch in activity then became apparent throughout late July, with rockfalls and pyroclastic flows initially coursing down Mosquito Ghaut. By 14 July rockfall activity was concentrated more to the west, directing pyroclastic flows down Gages valley. This shift is attributed to the stagnation of the 27 June 1997 lobe and the formation of the 14 July 1997 lobe directing an active headwall to the west. A further pulse in activity on 31 July was again marked by an abrupt increase in amplitude of tilt cycles in conjunction with intense hybrid earthquake swarms (Voight et al. 1999). This was associated with a growth spurt that produced a westerly directed dome collapse engulfing Gages valley and Plymouth Town in pyroclastic flow deposits on 3 August (Cole et al. 2002; Druitt et al. 2002). In a similar scenario to 25 June, the rapid influx of magma into the upper conduit had pushed out the entire shear lobe that was already actively growing to the west. This event produced a deep central crater in the dome with a breached, open scar facing to the west. The rapid removal of the 14 July 1997 lobe, caused by the collapse on 3 August, perturbed the magmatic system in the upper conduit and triggered a week-long series of repetitive Vulcanian explosions (Druitt et al. 2002). By mid-August, blocky lava was apparent within the crater and the emplacement of a westerly




Fig. 14. (a) 2 October 1996. Photo looking west from above the Tar River valley (see Fig. 13a). Castle Peak (CP) is the dark material in the left foreground surrounded by fresh, pale lava of the new dome, which appears as a rugged ridge defining the rim of the horseshoe-shaped collapse scar (CS) and explosion crater from 17 September 1996 collapse and explosive eruption. Small bun-shaped feature in the centre of this scar is the 1 October 1996 lobe (OL) extruded 15 days after explosive event had evacuated the upper conduit. Megaspine (M) of early July (see Fig. 10c) can be seen behind the central growth and Peak C on English's Crater is on far right. Trench in foreground is the main exit channel eroded by pyroclastic flow activity on 17 September 1996. (b) 2 October 1996. Close-up of new growth (1 October 1996 lobe, OL) seen in (a) showing smooth-topped massive lava (c. 20m thick) resting atop a mantle of rock debris. This feature, named informally 'the brain', gradually rose and infilled the 17 September 1996 explosion scar over the following month. Note the fumarolic activity around the perimeter of the fresh lobe.

directed 13 August 1997 lobe rapidly ensued. Elevated seismicity and extrusion rates (estimated at > 5 m 3 s -1) continued with gradual collapse of the headwall continuing to generate pyroclastic flows towards Plymouth. By late August, the western dome flanks had been rebuilt (Fig. 25b, c), with the height of the dome nearly reaching 1000m a.s.1.

In early September 1997, the broad headwall of the 14 August 1997 lobe was widening its span to generate rockfalls down the northern flanks, as well as down Gages valley. By 8 September, rockfall activity inundated the northern flanks whilst activity had shut down on the western flanks, indicating stagnation of the 14 August 1997 lobe. Lava blocks spilled off the active headwall of



130 R . B . WATTS E T AL.

Fig. 15. (a) 10 October 1996. View looking west from above NE dome flanks showing the 1 October 1996 lobe (OL) infilling the scar (ES) formed by the 17 September 1996 explosive eruption. Small scar (S, in foreground) was formed by small rockfalls tumbling down the eastern flanks of the dome. Note the rubbly carapace (a few metres thick) that characterized the flat surface of the 1 October 1996 lobe, c. 160m in diameter here. (b) 3 December 1996. View looking WNW at the dome from above southern rim of English's Crater (EC). The 1 October 1996 lobe (OL) growing in the 17 September 1996 explosion crater and collapse scar can be seen in the centre. The steep rim (S) of the spoon-shaped scar of 17 September is well seen here. ED marks the approximate zone of localized doming noted during volume surveys at this time and interpreted as a site of endogenous growth. ChP marks the summit of Chances Peak dome (summit height c. 909 m a.s.l.). Galway's Wall (GW) with fresh avalanche debris at its base is centre left, and a small remnant of Castle Peak (CP) is visible in central foreground.

the 8 September 1997 lobe now growing directly no r thward and debris completely filled Mosqu i to G h a u t at this time thus extend- ing the limits of Farrel l ' s Plain (Fig. 25c). The active lobe continued to spall away blocky lava onto the nor the rn dome flanks and down Tuitt 's Ghaut . The dome at this stage was vo luminous (c. 85 x 106 m3), with the nor thern flanks growing as a single lobe. On the morn ing of 21 September 1997, a swarm of hybrid earthquakes preceded a large dome collapse. A n est imated 14.3 x 106 m 3

of mater ial was funnelled down Tuitt 's Ghaut , forming an amphi thea t re -shaped scar in the dome, with a large crater open to the nor th (Figs 26a and 27a). This collapse was a further example of the rapid extrusion of a new shear lobe destabilizing the sector o f the volcano in which it was actively growing. A second series of repetitive Vulcanian explosions followed the collapse, due to rapid depres- surization of m a g m a in the main condui t (Drui t t et al. 2002). This mon th - long period p roduced only very minor changes to the overall




Fig. 16. (a) to (c) Maps highlighting extrusion of the 13 December 1996 lobe, the 25 December 1996 lobe, the 21 January 1997 lobe and the location of vigorous pyroclastic flow activity in mid- to late January 1997. Legend as in Figure 5.

dome morphology, although explosive activity reamed out a deep, funnel-shaped central crater, almost 300 m in diameter. The onset of renewed dome growth on 22 October involved the slow extrusion of another lobe to the north. This feature initially extruded vertically, with the back wall of the explosion crater acting as the shear

zone accommodating its growth. As it continued to rise, the 22 Octo- ber 1997 lobe gradually projected to the north, partly infilling the scar produced by the 21 September collapse (Figs 26b and 27b).

Growth of the 22 October 1997 lobe stagnated on 3 November and rockfall activity started to spill down the southern flanks of




Fig. 17. (a) 18 December 1996. View looking west from southern edge of Tar River valley marked by point Z on Figure 16a. Large structure in centre is the headwall of the 13 December 1996 lobe, around 150 m high and 200 m broad. VL marks the smooth, curved outer surface of the megaspine extruded in the early stages of this lobe. CS marks the scar rim of the September explosion crater and collapse scar; CP is the main spine of Castle Peak. Pale blocky talus in the foreground is rockfall debris produced by extrusion of the 13 December 1996 lobe. (b) 28 December 1996. Night-time view of glowing dome seen from the Whites area to the northeast, c. 2 km from the dome (Fig. 1). Incandescent blocks of the 25 December 1996 lobe are evident, spilling down the eastern flanks of the dome. Note the silhouetted twin-pronged spine of Castle Peak in the left foreground.

the dome, coincident with a period of intense shallow hybrid earthquakes. Three major collapses (c. 8 x 106m 3 total volume of deposits) then occurred on 4 and 6 November, each concentrated on the southern flanks and sending pyroclastic flows coursing down the White River valley. These collapses are attributed to the 22 October 1997 lobe dislocating (hence stagnating), with redirected extrusion to the south (Fig. 28). The bulldozing effect of the 4 November 1997 lobe instigated collapse of older lava on the southern dome flanks (the two collapses on 4 November) and collapse of freshly extruded blocky lava from the headwall of the 4 November

lobe itself (the 6 November collapse; Figs 29a and 30a). Rapid advance of this lobe infilled the collapse scar formed by these events, pushing a broad headwall of massive lava southwards. Lava continued issuing from this southerly directed shear zone for many weeks. Throughout this period, activity alternated between the extrusion of broad, curving spines at the back of the lobe or rockfalls spilling away from the summit headwall. This latter activity suggested a pulse of magma feeding directly into the core of the shear lobe and triggering rockfalls off the summit headwall (Fig. 9c). This alternation between spine extrusion and headwall break-up, on a




Fig. 18. Schematic diagram illustrating extrusion of the 13 December 1996 lobe. Section is along line X-Y on Figure 16a. (a) Dome configuration on 3 December 1996, with earthquake swarms indicating endogenous activity but no clear evidence for dome growth at surface. Magma within conduit continues to degas and crystallize until remobilized. (b) Dome configuration on 13 December 1996. A pulse of fresh magma has uplifted part of the 1 October 1996 lobe, resulting in the extrusion of a megaspine, the initial growth of the 13 December 1996 lobe. This structure is pushed to the SE, bulldozing through the dome flanks generating rockfalls down the Tar River valley. (e) Dome configuration on 19 December 1996. The 13 December 1996 lobe is overwhelmed and broken up by the extrusion of fresh, hot, blocky lava. Pyroclastic flows inundate the Tar River valley until the waning stages of the pulse. (d) Dome configuration on 28 December 1996. The shear zone feeding the 13 December 1996 lobe starts to plug with crystalline lava. A further pulse of magma emplaces more fluid lava at the summit of the dome (i.e. the 25 December 1996 lobe), rapidly spreading over the l October 1996 lobe and 13 December 1996 lobe and spilling lava blocks down the eastern flanks.

near-daily basis, suggested short-term fluctuations in the discharge rate and alternating pulses of more viscous and less viscous magma.

An immense peak was gradually constructed that straddled the Galway's Wall area (Figs 29b and 30b). By 21 December, the summit of the 4 November 1997 lobe was c. 1030m high and the loading of this structure was undermining the strength of the Galway's Soufri~re area, which was progressively buried. During a period of intense hybrid seismicity on 26 December, a debris avalanche and major dome collapse (known as the 26 December

1997 or 'Boxing Day' collapse) removed the entire 4 November 1997 lobe (c. 55 x 106 m 3 non-DRE of dome rock and talus) that had grown since early November (Sparks et al. 2002; Voight et al. 2002). Remarkably, the other sectors of the dome were unaffected by this major flank failure and dome collapse.

Vigorous extrusion of blocky lava began to infill the resulting collapse scar immediately (Fig. 31a). Renewed development of a southerly directed lobe, the 27 December 1997 lobe became apparent, and within two months, the southern flanks were rebuilt



134 R. B. WATTS ET AL.

Fig. 19. Cartoon illustrating the development of the 25 December 1996 lobe (see Fig. 16b). Sketches on the left represent a NW-SE cross-section through the dome; sketches on the right represent plan views at the same time. (a) 28 December 1996. The 13 December 1996 lobe has stagnated. The 25 December 1996 lobe punches vertically through the 1 October 1996 lobe and proceeds to spread symmetrically across the summit, overwhelming the 13 December 1996 lobe. (b) 5 January 1997. Lava extrusion becomes more directed, guiding curved slabs to the east and overwhelming the 13 December 1996 lobe and spilling lava blocks down the eastern flanks. The western edge of the 25 December 1996 lobe has become abandoned from the rest of the lobe. (e) 17 January 1997. Oversteepening of the SE dome flanks triggers a dome collapse on 16 January, and ensuing blocky lava infills the horseshoe-shaped scar. Lava extrusion is focused within the scar and directed by a southeasterly shear lobe as a large sector of the 25 December 1996 lobe is abandoned.

(Fig. 30b), although the growth rate had noticeably waned as this period progressed. Development of this Type 1 shear lobe occurred as alternating spine extrusion and rockfalls off the summit headwall suggesting endogenous activity, a similar pattern to that observed in November and December 1997. By late February, several large spines had extruded prior to the formation, in early March, of a prominent 50-m-high spine (informally termed the Galway's Spine) that towered atop the 27 December 1997 lobe (Figs 31b and 32a). Coincident with extrusion of this feature was a marked reduc- tion in rockfall activity, seismicity and ground deformation; all of

these factors signalled the cessation of the first episode of dome growth on about 10 March.

Stage VII." 11 March 1998 to mid-November 1999

A full account of this 20-month interim period between the two phases of dome growth is presented by Norton et al. (2002). Despite no extrusion of fresh lava at the surface throughout this period,




Fig. 20. (a) 22 January 1997. View of same area as Figure 17a but closer and one month later on, following a period of vigorous blocky growth and subsequent dome collapse. Flower-shaped structure (with radial cracks) is extrusion of massive lava (c. 30 m high) of the 21 January 1997 lobe sitting within the 20 January collapse scar (CS). Note the spine of Castle Peak (CP) in right foreground and the nearly vertical fractures within this spine. (b) 25 January 1997. Similar view taken three days later than (a), with the same features marked. The 21 January 1997 lobe developed into a pile of curvilinear, massive lava blocks (typically c. 5 m in size) which has continued to infill and engulf the collapse scar (see Fig. 16c). This blocky morphology was characteristic of a Type 2 shear lobe formed during periods of relatively high discharge rate (>5 m 3 s-i). Many of these blocks tumbled down the SE dome flanks (in foreground) to generate moderate-sized pyroclastic flows down the Tar River valley.

there were still intermittent periods of increased activity, notably the 3 July 1998 dome collapse (Fig. 32b). This large-volume collapse followed three months of quiescence with only rare rockfalls, small pyroclastic flows and subdued seismicity. The collapse involved 20-25 • 106 m 3 of lava and formation of a 200-m-deep horseshoe- shaped canyon through the dome on its southeastern flank. There was no seismic precursor to this event, which is believed to have been predominantly gravitationally influenced, although heavy rainfall may have been a contributory factor. The collapse was

initially focused in an area weakened by continuous intense fumarolic activity. The southeastern flank was also a steep ridge of loose, blocky lava that formed the eastern rim of the scar from the collapse of 26 December 1997, and was therefore prone to instability. Following the collapse there occurred sporadic degassing and small explosions from a crater at the central base of the scar. A smaller collapse on 12 November 1998 removed c. 3 x 106 m 3 of lava from the western dome flanks. The collapse eroded away part of the western dome flanks and joined the 3 July 1998 scar forming a deep



136 R . B . WATTS ET AL.

Fig. 21. (a) 1 April 1997. View looking west above Tar River valley showing the conical form of the dome that developed throughout March 1997. As blocks tumbled eastwards off the active area they eroded deep rockfall channels and chutes into the talus of the lower dome flanks. Note that Castle Peak has been completely overwhelmed by fresh lava. Four days previously, the easterly directed 21 January lobe (JL) stagnated and the 27 March 1997 lobe (EL) started to grow towards Galway's Wall (G). Summit height of dome c. 970 m a.s.l. (b) 6 April 1997, from near the same location as (a). This photo highlights the dramatic contrast between the conical-shaped dome from growth in February and March 1997 (JL) and the initial, smooth appearance of the active 27 March 1997 lobe (EL) sliding out southwards (to the left) from the summit area. This also highlights the contrast in morphology of a Type 1 shear lobe in the early stages (EL), and the later stages (JL) of lobe development (see Fig. 9b, c). In this view, the 27 March 1997 lobe was around 200 m long, and 150 m wide, with a c. 150 m high headwall that generated dome collapses down the White River valley as it grew. The arrow indicates motion. Note the gas plume emerging from along the boundary between the two parts of the dome.

corr idor-shaped gorge t rending E S E - W N W through the entire dome. A further collapse on 20 July 1999, this time directed to the east, removed c. 5 x 106 m 3 of lava, mainly off the 27 December 1997 lobe. This event excavated a smaller canyon th rough the nor thern flanks that l inked up with the main gorge (Nor ton et al. 2002). Sporadic explosions and mino r gravitat ional collapses con- t inued to occur until the onset of renewed dome growth in mid- N o v e m b e r 1999 (onset of Stage VIII; Table 1).

Discussion

Petrological and rheological variations

Lava dome extrusion is p rofoundly influenced by magma rheology. We begin our discussion of the observat ions of the Soufri~re Hills dome evolution by summariz ing the petrological characteristics of the andesite. These features provide major constraints on the




Fig. 22. (a) and (b) Maps illustrating the dome configuration in March/April 1997 showing how extrusion of the 27 March 1997 lobe affected the morphology of the dome. Legend as in Figure 5. (c) 6 April 1997. Close-up view of the 27 March 1997 lobe taken from point X on (b). Note the smooth, upper, semi-cylindrical surface of this lobe (BW), and the prominent striations trending parallel to the direction of shear movement, with deep, cross-cutting fractures also evident. Visual observations showed this feature to be sliding out in a stick-slip manner at an estimated 25-30 m day -1 . This lobe was c. 200m long (from left to right) and c. 150 m wide with a e. 150 m high, steep headwall. Rockfalls were continually spalling off the near-vertical headwall (H) at the leading edge of the lobe (in the left foreground) as a result of intermittent endogenous/exogenous activity. This lobe represents the only example throughout the g1995-1998 period where the smooth, upper surface remained intact during growth, rather than breaking up into large, curving slabs. JL is stagnated lava from the 21 January 1997 lobe.

rheological variations and on the principal factors that control dome extrusion.

Throughout the 1995-1999 period, the dome andesite was sampled by collection of blocks from dome-collapse and fountain- collapse pyroclastic flow deposits. Petrological work (e.g. Devine et al. 1998; Murphy et al. 2000) highlighted only a minor variation in

bulk composition (SiO2 58.5-62.4%) and phenocryst/micropheno- cryst content (55-65%). The eruption during 1995-1999 is interpreted to have been driven by an open-system magma body fuelled by the influx of mafic magma into a very crystal-rich magma body resident at shallow crustal levels. Both the thermal input and pressurization due to mafic magma influx (reflected in the 1992-1995



138 R.B. WATTS E T AL.

Fig. 23. (a) 6 April 1997. View looking NW from White River valley area highlighting the destruction of Galway's Wall by growth of the 27 March 1997 lobe. EL marks the steep headwall (c. 150 m high) of the lobe in which arcuate cooling fractures and zones of intensely sheared lava are evident. WF is the western dome flanks and S the highest point on the dome (c. 970 m a.s.1.), composed of lava extruded in February and March 1997 (21 January lobe) and now inactive. Note the deep erosional gully in Galway's Wall formed by numerous rockfalls and small pyroclastic flows sourced from the 27 March 1997 lobe. (b) Same view as (a) but taken in mid-May 1997, showing the construction of a broad talus of lava blocks, slabs and small spines that has completely buried the former Galway's Wall and the early plug-type extrusion feature of the 27 March 1997 lobe seen in (a).

seismic crisis and seismic crises in the previous 100 years) was sufficient to remobilize the source region to form a crystal-rich andesite magma. The samples contain 35-45 vol% phenocrysts and c. 20 vol% microphenocrysts within a microlite-bearing high-silica- rhyolitic glass matrix. The main difference between samples has been the degree of crystallinity and texture of the groundmass, both factors affected by the magma discharge rate. Between November 1995 and mid-February 1996, the erupted lava had a highly crystalline groundmass with only 5-15% residual rhyolitic glass and typically extruded as large spines. This early phase of activity is interpreted to be the extrusion of degassed lava infilling the conduit from previous injections of magma that triggered the seismic crises. This is supported by the identification of amphiboles with heterogeneous hydrogen isotope compositions in samples from this period

(Harford & Sparks 2001). Samples from periods of rapid dome growth from August 1996 to March 1998 have tended to include higher glass contents (up to 30%), although the glass content range is wide (5-30%). The major pyroclastic flows sampled the deep interior of the dome to depths of 100-200m, well away from the influence of surface cooling. These samples still exhibit extensive groundmass crystallization that are attributed to degassing rather than cooling (Sparks 1997).

The ascent of andesitic magma from the chamber to the near- surface environment has been modelled (e.g. Sparks 1997; Melnik & Sparks 1999, 2002) and changes in rheological properties are attributed to two intimately linked mechanisms. Gas exsolution due to decompression during slow ascent causes a large increase in the melt phase viscosity (Dingwell et al. 1996). Degassing also triggers




Fig. 24. Schematic diagram illustrating the dramatic switch in activity that was experienced in mid-May 1997. Sketches on the left represent a N-S cross-section through the dome whilst sketches on the right represent plan views of the dome at the same time.

crystallization in response to the consequent undercooling of the melt phase and this process also increases magma viscosity. The efficacy of both mechanisms increases during ascent and reaches a peak in the uppermost several hundred metres of the conduit, where Melnik & Sparks (2002) predict a zone of large overpressures. Degassing-induced crystallization has already been noted by many workers (e.g. Sparks & Pinkerton 1978; Lipman et al. 1985; Wolf & Eichelberger 1997; Gardner et al. 1998; Blundy & Cashman 2001) and is invoked for the Soufri~re Hills magma (Sparks et al. 2000). However, the potency of this mechanism in andesitic dome-forming eruptions had not been fully appreciated. Throughout most stages of dome growth in the Soufri~re Hills eruption, a persistent gas plume was seen emerging from around the dome summit, indicating effective degassing of the magmatic system. The consequence of degassing was to cause a profound rheological stiffening of the magma so that the lava was better characterized as hot, crystalline material with considerable strength (Sparks et al. 2000).

The rheological properties of the Soufri6re Hills andesitic magma can be constrained from petrological observations and uniaxial loading experiments on dome samples at high temperature (Sparks et al. 2000). Petrological estimates of magma properties in the magma chamber suggest temperatures of c. 860~ with c. 5% H20 dissolved in the rhyolitic melt phase; thus a viscosity of 7 x 106 Pa s is estimated, based on experimental results (Dingwell et al. 1996; Pinkerton & Stevenson 1992). Fully degassed and highly crystalline dome samples show highly non-linear deformation behaviour under uniaxial loadings of 9-26 MPa at temperatures of c. 990~ (Sparks et al. 2000). Viscosities in an initial period of steady deformation are of the order of 1014 Pa s, but apparent viscosities decrease rapidly to c. 1011 Pa s just prior to failure along shear zones.

The Soufri6re Hills andesitic magma was already rich in phenocrysts and microphenocrysts in the magma chamber prior to eruption. During periods of slow ascent, degassing-induced




80500

t t t

t

81400 BUbUU

>

> > >

~ i ) ii i ~ ~i!' �9 ~ " " ' , 7 / , ~ ,

19 J U L Y 1997

s s

A ~ A

! i il ! ,

C

i ,

:~!-i:!! ::::::ii) i::

D o m e V o l u m e = ,,-76.0 x 106 m 3 13 A U G U S T 1997 3 -1

E x t r u s i o n R a t e = ~ 3 - 4 m s

,>

o

47600

/

1

46600

D o m e V o l u m e = - 80 .0 x 108 m 3 3 ~1

E x t r u s i o n R a t e = ~ 4 - 6 m s

80500

A A A A

A A J A

A A A A i

81400 47600

. . . . ~ i ~ . . . . . . . . ~ o " ,

+

31 AUGUST 1997 Dome Volume = -82.0 x 106 3 -1

Extrusion Rate = - 4 - 6 m s

46600

KEY

13Augus, 1997,obe

I I A A

A A

' ' 14 Ju 1997 lobe A

li-" ~ ~' I 17 May 1997 lobe

:i!~:! 27 March 1997 lobe

I I January 1997 lobe

Pre-Sept 1996 lava

Fig. 25. Sequence of maps illustrating the growth of the dome in the aftermath of the 25 June 1997 collapse and highlighting the development and subsequent infilling of the explosion crater throughout August 1997. Legend as in Figure 5.

crystallization of magma in the upper conduit reached the threshold time for microlite crystallization to the extent that the critical when crystals formed a touching framework. At this stage the threshold was not attained and magma could extrude in a more fluid- magma transformed from a Newtonian fluid to a much more viscous like manner. Thus, during a period of fluctuating discharge rates, the non-Newtonian fluid with mechanical strength (Lejeune & Richet flow of crystalline magma may switch between a fluid nature and 1995). In contrast, a sufficiently fast magma ascent rate reduced the that of a near-solid, non-Newtonian nature. Such an oscillating




Fig. 26. Maps showing the configuration of the 21 September 1997 collapse crater and subsequent growth of the 22 October 1997 lobe from 22 October to 3 November 1997. Legend as in Figure 5.

state is predicted by models of the non-linear dynamics of conduit flow (Denlinger & Hoblitt 1999; Melnik & Sparks 1999, 2002; Voight et al. 1999; Wylie et al. 1999). A consequence of this variation is heterogeneous deformation during lava emplacement, with the formation of spines and megaspines at slow rates (for non- Newtonian lava) and the development of shear lobes emplacing more ductile lava as large blocks and smooth slabs or pancake- forms at faster rates.

Morphologic variation

A spectrum of extrusive features was observed throughout the 1995-1998 episode of dome growth, and together these structures can be considered to represent a morphologic continuum (Fig. 33). Each structure was essentially composed of the same components (i.e. a smooth, semi-cylindrical backwall and a steep, blocky headwall), and their growth bounded along shear faults was accommodated in the same manner, with the exception of pancake forms at more vigorous rates. Periods of growth varied widely, however, with spine and megaspine formation occurring in a few days, whilst individual shear lobe evolution operated over several weeks and months. The main difference between these structures related to their size and there appeared to be a link between the formation of each structure and the level of eruptive activity (Fig. 34). Only occasionally did the lava morphology and behaviour suggest a more fluid-like emplacement, with axisymmetric lateral spreading

of pancake lobes at the summit area during periods of more vigorous discharge rates (Fig. 33). The activity of late December 1996 provided the best example of this, with emplacement of the 25 December 1996 lobe occurring at a time when seismic tremor was commonly experienced. A similar relationship, linking growth behaviour to discharge rate, has also been determined in experiments using a Bingham plastic analogue (Griffiths & Fink 1997; Fink & Griffiths 1998). These laboratory experiments reproduced many of the structures observed in the Soufri~re Hills eruption, although their formation was attributed primarily to variations in viscosity and the thickness of the cooled dome carapace, whereas here we relate the variations to the combined effects of discharge rate, cooling and degassing and related changes in the rheological properties of the magma.

On considering the morphologic continuum, a gradual transition from highly crystalline structures (at low discharge rates) to those exhibiting more fluid-like features (at higher discharge rates) exists. By far the most predominant style of activity was the formation of shear lobes at moderate discharge rates, although two distinct classifications have been observed. In Type 1 shear lobes (Fig. 33), growth develops a stable structure that may grow over many weeks and months, constructing a lobe with a core of massive lava and steep talus flanks from semi-continuous rockfall activity. Such activity is prevalent during periods of long-term average extrusion rates of 2-5 m 3 s -1. In contrast, during periods of fluctuating extrusion rates from <1 m 3 s -1 to >5m 3 s -1 (e.g. July to August 1996), emplacement of megaspines would be rapidly followed by



142 R.B. WATTS E T AL.

Fig. 27. (a) 28 September 1997. View looking south from the north (near Harris Village), showing amphitheatre in the dome following the 21 September 1997 collapse. Breach in the dome faces into the head of Tuitt's Ghaut (T). Y represents the headwall of the stagnated 17 May 1997 lobe and Z the headwall of the stagnated 13 August 1997 lobe (see Fig. 26a). The main crater is c. 300 m in diameter. (b) 6 November 1997. High aerial view looking towards southwest at the northern flanks of the dome showing the 22 October 1997 lobe, with headwall of massive lava sitting within the breached part of the 21 September 1997 crater. X marks the smooth, curving upper surface of this lobe. T, Y and Z represent the same features marked in (a). Activity had completely stagnated on the northern flanks, with major collapses affecting the southern flanks of the dome.

Type 2 shear lobes of blocky lava (Fig. 33). At such periods, the rapid sequential emplacement of these very different structures commonly triggered dome collapses with major pyroclastic flow activity. As the eruption progressed, a steadier discharge rate became established, promoting the formation of Type 1 shear lobes. As a result, long periods of lobe construction, with associated accumulation of rockfall debris, would occur with only minor pyroclastic flow activity (e.g. January to March 1998). This latter style of dome growth has been particularly predominant in the second phase of dome growth that commenced in November 1999.

Another factor to consider when explaining the difference in activity between the earlier and later stages of the eruption may be the former presence of the Castle Peak dome and its gradual destruction and complete burial by early February 1997. In the early eruptive stages, emplacement was partly controlled by the topography of Castle Peak. The directed explosion of 17 September 1996 was also an important moment in the eruption. Not only was it the first magmatic explosive activity, it completely exposed the head of the main conduit to surface conditions, as well as widening the uppermost parts of the conduit. Prior to this event, magma had




Fig. 28. Schematic diagram illustrating the hypothesized growth of the dome during extrusion of the 22 October 1997 lobe and the 4 November 1997 lobe. The cross-sectional view is along the line X-Y on Figure 26a. (a) 23 October 1997 (see Fig. 26a). Renewed slow dome growth in the central crater following a month-long period of Vulcanian explosions. (b) 29 October 1997. Gradual extrusion of 22 October lobe, initially guided vertically by backwall of crater and subsequently projecting to the north, infilling the breached part of the crater and directing rockfalls down Tuitt's Ghaut. As the 22 October 1997 lobe extrudes subhorizontally away from the crater, two small hummocks of lava extrude at the central summit area (see Fig. 26b). (e) 4 November 1997. Growth of the 22 October 1997 lobe and central hummocks has stopped while the southern dome flanks are destabilized by intrusion of the 4 November 1997 lobe generating a dome collapse down the White River valley. (d) 6 November 1997. The southern dome flanks have been completely destroyed, whilst all the other flanks of the dome are unaffected. As the 4 November 1997 lobe continues to extrude along a southerly directed shear fault, fresh lava breaks off the lobe's leading edge causing a further collapse directed down the White River valley. (e) 21 December 1997. Growth of the 4 November 1997 lobe has continued for almost two months, constructing a large peak which now straddles the former Galway's Wall.

to force its way through the fractured body of the Castle Peak dome before extruding.

Structural control on emplacement

The dominant mode of emplacement following 17 September 1996 was through the formation of shear lobes bounded by large arcuate faults. During emplacement, the lava extruded sporadically in a stick-slip manner along curved fault structures, which are interpreted as shear faults that are sourced from the sides of the conduit. Similar features have been produced in analogue experiments by Donnadieu & Merle (1998) investigating the deformation of a volcanic edifice through forced intrusion of viscous magma. As indentation proceeded, asymmetric deformation generated a curved shear fault from the base to the outer edge of the edifice, with the fault controlling the directed emplacement of the magma analogue. This mechanism was postulated by Donnadieu & Merle (1998) to explain the cryptodome intrusion in the build-up to the Plinian eruption of Mount St Helens in May 1980.

Observations of the Soufri6re Hills dome indicate that a similar mechanism may be responsible for the growth and development of individual lobes during construction and destruction of the dome. Shear faults have not been recognized in lava domes before. An explanation for this may be the poor preservation potential of the shear surfaces. In the Soufri+re Hills dome, a lobe was commonly broken up in the late stages of emplacement or buried by blocks from a later eruptive episode. The structure of a shear lobe is sometimes preserved and, at the time of writing (October 2000),

the current configuration of the Soufri~re Hills dome still exhibits part of the smooth backwall of the 4 November 1997 lobe extruded in October 1997. Domes at other volcanoes also have preserved examples, such as a 55-m-long, 25-m-wide striated lobe emplaced near the summit of the Chinois Dome that formed in the 1929-1932 eruption of Mont Pel~e, Martinique. In some cases, a single shear lobe may form the entire preserved part of a dome; the 34-ka Perches dome in the Soufri~re Hills complex on Montserrat and the Gros Piton on St. Lucia are both examples of near-vertical shear lobes.

Growth stages and cycles

During the eruption the dome has increased in volume, notwith- standing the counteracting effects of collapses. Growth has also been characterized by cyclic patterns, with repeated switches in the direction of lobe extrusion and pulsations in discharge rate. Here we discuss the interplay of individual pulses of lobe extrusion with the overall construction of the dome and the nature of the cyclic growth patterns.

The early stages of the first episode of dome growth involved the gradual destruction and burial of the previous construct, the Castle Peak dome. As eruptive vigour increased, the directed extrusion of megaspines and subhorizontal shear lobes occurred in a non-systematic radial pattern around the central conduit. These structures armoured the lower dome flanks, acting as a foundation to the large subvertical shear lobes that dominated extrusion in the later eruptive stages. Upon stagnation, the broad headwall of each




Fig. 29. Maps illustrating the collapse resulting from growth of the 4 November 1997 lobe and the extent of regrowth following this event.-Legend as in Figure 5.

lobe would form the steep upper ramparts on a sector of the volcano. This process of shear-lobe growth and stagnation continued around all sectors of the dome, so as eventually to construct a central depression, which was further modified by explosive activity in August, September and October 1997. This depression developed as a result of this repeated lobe formation, with each lobe directed away from the central conduit. In effect, the remnants of the smooth outer lobe surfaces had stagnated and merged together to form the inner walls of the central depression. In association with this process, each of the headwalls of the lobes stagnated and merged together to form the steep upper flanks of the dome. As extrusion of individual lobes constructed steep flanks, extrusion would either continue in the same direction, partly stagnate and widen the headwall to one side of the lobe, or completely stagnate and switch the focus of activity to another sector.

Major switches of dome growth direction can, in several cases, be clearly linked to cyclic behaviour of the volcano. Throughout most of the eruption, and particularly in the latter half of the 1995- 1998 period of dome growth, a distinct five to seven-week cycle of activity was recognized, which was intimately linked to hybrid seismicity and a pronounced increase and decrease in the period of tilt cycles (Voight et al. 1999). Typically, the onset of a cycle was marked by a period of intense hybrid seismicity that commonly resulted in a dome collapse event and a pronounced increase in discharge rate. In the weeks that followed, the amplitude and period of tilt cycles decreased and increased respectively and seismicity declined. Aseismic growth of the active lobe infilled the collapse scar, initially as rapidly emplaced blocky lava and small spines that

developed into a broad shear lobe with oversteepened flanks. As predominantly aseismic growth progressed, the discharge rate would gradually wane over several weeks, until dropping below a critical threshold. At this point, degassing-induced crystallization could operate more effectively to form a near-solid plug in the upper conduit, contributing to the retardation of surface extrusion. The onset of a new cycle is then attributed to a pulse of fresh, less viscous magma, rising up from the deep source. The new magma would reach the base of the consolidated plug and the resistance to upward flow marked the onset of hybrid seismicity. Pressure build-up beneath the plug would continue over a few hours to several days, as highlighted by intensifying swarms of hybrid earthquakes until reaching a critical limit that pushed the plug into the dome. The discharge rate rapidly increased as less viscous magma filled the conduit and the plug material was pushed out of the way. The rapid increase in flow rate and pressurized conditions in the upper conduit were, in many cases, enough to rapidly destabilize a lobe headwall and trigger a major dome collapse. The fatal 25 June 1997 dome collapse illustrates a typical growth cycle and, in the weeks leading up to the event, aseismic growth of a northerly directed lobe had primed the northern flanks for a collapse (Fig. 35a). In the weeks following the collapse, rapid redevelopment of a fresh northerly lobe was observed (Fig. 35b), with hybrid seismicity disappearing, thus initiating the next cycle (Fig. 35c). In this interpretation, the fluctuations in discharge rate observed during each cycle were a consequence of pulsations of fresh, low-viscosity and probably gas-rich magma released from the deep source with hybrid seismicity being a signature for the extrusion of a near-solid




Fig. 30. (a) 8 November 1997. View looking north directly at the headwall (NL) of the 4 November 1997 lobe (c. 150 m high). Dotted line marks the original southern dome flanks immediately prior to extrusion of the 4 November 1997 lobe. Note also its curved upper edge (BW) and the vigorous ash-venting emanating from the rear of this lobe. (b) 27 February 1998. This excellent view of the southern flanks, taken high above the White River valley, highlights the dome configuration almost as it was immediately prior to the 26 December 1997 collapse. This shot was taken many weeks following this collapse; however, subsequent growth infilled the collapse scar (BD) and rebuilt the southern flanks in a near-identical manner. Note the characteristic rockfall chutes leading down from the blocky summit and eroding into the talus. Pyroclastic flow deposits from the 21 September 1997 collapse can be seen in the background, with Chances Peak (ChP) and Galway's Mountain (GM) dome also in view.

plug into the dome, possibly by stick-slip movement along the walls of the upper conduit (Denlinger & Hoblitt 1999; Voight et al. 1999; Wylie et al. 1999).

Models o f dome growth

The observations of the growth structures and morphological development of the Soufri6re Hills lava dome are now discussed in

the context of concepts and models of dome growth. Two main concepts have been developed to explain dome growth. First, the role of surface cooling with formation of a resisting crust have been explored by Fink & Griffiths (1992) in a series of laboratory experiments and by scaling analyses of force balances (summarized in Griffiths 2000). Second, the linked roles of gas exsolution and degassing-induced crystallization have been considered in textural studies (Cashman 1992; Cashman & Blundy 2000; Hammer et al. 2000) and theoretical models of conduit flow during dome extrusion




Fig. 31. Maps illustrating the dome configuration in the aftermath of the 26 December 1997 collapse and the focus of regrowth following this event until the cessation of growth around l0 March 1998. Legend as in Figure 5.

(Sparks 1997; Melnik & Sparks 2002). In both concepts, rheological stiffening and the onset of non-Newtonian rheology with the development of a yield strength are important. Here we discuss the relative roles of cooling and degassing-induced crystallization in controlling the dynamics of the Soufri6re Hills lava dome and morphological features. This discussion develops that by Sparks et al. (2000).

In Griffiths (2000) similar phenomena and range of dome morphologies to those reported here are attributed to development of a cooled crust. Fink & Griffiths (1998), for example, demonstrate a similar spectrum of structures to that observed at Soufri&e Hills, from spiny lobate domes to smooth-spreading pancake morphologies, by changing a dimensionless parameter �9 which reflects the relative importance of heat advection (and therefore discharge rate) and heat loss through a cooling crust. Here we invoke a different view that this morphological spectrum is a consequence of different amounts of both degassing-induced crystallization and cooling. For lava extruded at the lowest rates, the slowly rising magma had time to lose more gas and crystallize efficiently, such that solidi- fication was largely completed (90-95%) within the upper conduit, resulting in the emplacement of a spine or a megaspine. Cooling played a negligible role in this case. For faster-rising magma, lesser amounts of microlite crystallization occurred during ascent, promoting the formation of shear lobes. At the fastest ascent rates, lava extruded in a more fluid-like manner and a pancake-type morphology resulted. In this case, emplacement could be controlled both by degassing-induced crystallization and external cooling.

Another argument for the importance of a cooling crust put forward by Griffiths (2000) is that many lava domes show an increase in height with time, as in the case of Soufribre Hills (Fig. 3a). This is attributed to growth of a crust of increasing thickness. By comparing the forces driving lateral spreading of a dome with the yield strength of a cooled crust, a scaling result is obtained whereby the height is proportional to time to the power of 0.25. Griffiths (2000) shows that several domes indeed exhibit a power-law depend- ence, with a power approaching 0.25. Such an analysis is prob- lematic for Soufri6re Hills because the dome collapsed many times (see Fig. 3a). It is more meaningful to take a single episode of growth. For the period mid-February to August 1996, height data yield a best-fit power law with the exponent of 0.36. In another example, the period 1 October 1996 to 5 November 1996, the height data give a power exponent of 0.44. In neither case are these values close to that predicted by the cooling-crust model of Griffiths (2000).

There are several problems with applying the cooling-crust model to Soufri6re Hills. Firstly, the overall increase in height with time does not result from the growth of a single entity with thickening and strengthening of a cooled crust. The dome collapsed and new lobes extruded many times, so the scaling analysis and the growth of a single, cooling dome structure cannot be applied to the whole dome growth of 1995-1998. Secondly, the cooling-crust model is a static one, whereas Stasiuk & Jaupart (1997) and Melnik & Sparks (2002) show that there are important dynamic controls on dome height. The increase in height with time is attributed to increasing magma chamber pressure by Melnik & Sparks (2002),




Fig. 32. (a) May 1998. View looking NW from South Soufri+re Hills at the southern flanks of the dome, composed entirely of the 27 December 1997 lobe. Galway's Spine (GS) on top of the blocky flanks is 50 m high and 45 m wide and its extrusion in early March signalled the cessation of dome growth. Chances Peak (ChP) is to the left, and eastern flanks of the dome are on the far right. The scar rim of the 26 December 1997 collapse is marked BD. (b) View looking NW from above the southern rim of English's Crater in early February 1999. The entire SE dome flanks have collapsed away exposing a vertical section (c. 200 m high) through the hummocks extruded as part of the 22 October 1997 lobe (OL). BD marks the same location as in (a), NL marks the remaining part of the 27 December 1997 lobe and the summit of the new dome at 977m a.s.1.

who consider it to be unrelated to surface cooling. Thirdly, consideration of height versus time for the 1 October 1996 lobe gives a power law with an exponent of 0.44, significantly larger than the 0.25 value expected by a cooling model. Likewise, the height versus time data can also be analysed by a dynamical model (Melnik & Sparks 2002) without invoking cooling. There is a more general problem in that several different models can give quite similar power-law behaviour, so that finding a power law exponent of 0.25 is not sufficient to demonstrate that cooling is the dominant effect. Indeed

a simple dynamical model, where discharge rate is a linear function of dome height, gives a height versus time relationship quite close to a power law with exponent of 0.25. Fourthly, the cooling model requires very high crustal strengths (> 108 Pa; Griffiths 2000), which are hard to reconcile with laboratory data on rock strengths. For example, geotechnical measurements (Voight et al. 2002) and experiments simulating explosions (Alidibirov & Dingwell 2000) indicate strengths of about 106 Pa for dome samples. Fracturing of the cooled crust and development of tensional cooling stresses further




E M P L A C E M E N T F E A T U R E + A S S O C I A T E D S E I S M I C l T Y

A. NEAR-VERTICAL SPINE

-growth over 2-3 days -coincident with periodic hybrid earthquake swarms

B. WHALEBACK STRUCTURES

-growth over 4-5 days -coincident with repetitive hybrid earthquake swarms

C. MEGASPINE

-emplaced over 2-3 days

-generally aseismic growth occasionally with hybrid seismicity

D. SHEAR LOBE-TYPE 1

-growth (broad spines) over many weeks to months, in form of intermittent endogenous + exogenous pulses

-generally aseismic growth, often intense hybrid earthquake swarms prior to lobe collapse

E. SHEAR LOBE-TYPE 2

-growth (blocky lava) over days to weeks, also with endogenous pulses

-growth and collapse often coincident with repetitive hybrid earthquake swarms + tremor

F. PANCAKE LOBE

-growth over 4-6 days

-emplacement coincident with repetitive hybrid earthquake swarms + tremor

G. EXPLOSIONS

-commonly occur following rapid, large dome collapses

-see Druitt et al. (2002)

I I I I

20m;

Om$ 7,

I / I I

i i ..................................., --";Om I I I , l l

~@,'. ~ ...................... ~.? '(.Q" ~ " / _ " " . ~ ,

/ ~ massive \'~'~, / / "... core -' \ / / .............................. iY50ml \

, ~

I I

t l l ~ /

A V E R A G E D I S C H A R G E R A T E , m ~ S "1

0 . 5

1 .0

2 . 0

5 . 0

7 . 0

9 . 0

F- Z iii L~ CO <

(.9 Z rY

a

l- z < Z

0 a

>, I (.9 Z

(.3

-~ I I

--i _J

or)

=> I I

Y I c'l

Z

I LU

~ I

Z

< LU n,'

_J o 0 0 ii

o F- 0 iii ii ii iii

Fig. 33. Variation in the type of structure emplaced in relation to the average discharge rate, and the relative roles of degassing-induced crystallization and cooling. Note that the boundaries between eruption rates are arbitrarily defined.




Fig. 34. Graph showing the structures extruded throughout the eruption. This highlights a relationship between the type of structure extruded and average discharge rate suggesting that changes in dome morphology may be a crude proxy for estimating discharge rate during the eruption.

14

'7 12

03 E 10

0~ 8 t~ K , . .

6 K , . .

t~ r 4 r o0

~

s 2

0 0r

I ~ Type 1 shear lobes + blocky lava

pancake lobe explosions �9

near-verticalspines : ~ - megaspines + Type 1 + 2 shear lodes

13 December whaleback structures

!

200

1996

4 November 1997 lobe

\ 1997 lobe ** * I~

\ �9 .

" " 7" �9 �9 $ �9 * 22 October 199 lobe

�9 * * * *~ "~ "27 March 1997 lobe *t

4oo I ' 600 800

1997 1998 Time (days s ince start of eruption)

Galway's Spine + growth cessation

1000

weaken a cooled crust, an effect inconsistent with the high strength indicated from the cooling-crust model. While we concur that cooling can have a major role in many instances of lava emplacement, we suggest that it only becomes a factor during the most rapid periods of dome growth at Soufri6re Hills.

Comparison with other documented dome eruptions

The growth of the Mount St Helens dacitic dome between 1980 and 1986 allowed the first intense scientific monitoring campaign of a lava dome this century. Swanson & Holcomb (1990) detailed the complex growth history of the dome and described distinct patterns throughout its duration. Episodic lobe extrusion generally lasting several days was the predominant style; however, a one-year period of continuous endogenous growth was also observed. Three distinct periods of episodic growth showed a linear pattern in relation to the long-term growth rate. Each extrusion was generally preceded by a one to three week period of endogenous activity that often faulted and fractured the dome carapace and caused deformation of the crater floor. Between growth episodes, the dome would slowly spread laterally and subside as its hot, ductile core deformed due to gravitationally induced stress. The generally consistent shape of the dome for most of the eruption suggested a possible controlling mechanism. Iverson (1990) modelled the dome morphology as a core of relatively low strength enclosed by a strong, brittle carapace. This cooled, outer surface skin and the net effective viscosity of the lava were believed to play an important role in determining the location and manner of lobe emplacement on the dome.

Variations in the surface texture of the lava throughout the 1980-1986 eruption also highlighted a relationship between the underlying slope and the water content of the lava (Anderson & Fink 1990). Several types of lobes were distinguished, each type characterized by varying degrees of smooth and scoriaceous surface textures. Growth was often focused along a large, smooth fracture known as a crease structure that formed predominantly on shallow slopes. A progressive increase in the formation of smooth-surfaced lava observed during the eruption was attributed to more effective degassing of the magma during ascent and emplacement. The features formed on the Mount St Helens dacite dome and mechanisms responsible for their formation show only limited similarity to those observed at the Soufri6re Hills andesite dome.

The more recent 1991-1995 dome-forming eruption of Mount Unzen, Japan, provides another comparison (Nakada et al. 1999). This dome grew on a steep and unstable slope such that the dome sporadically partially collapsed generating pyroclastic flows. During this period of dome growth, magma extrusion was near-continuous, involving the exogenous extrusion of 13 distinct lobes. Initial growth

involved spine formation and subsequent collapse into blocks; however, later lobes commonly exhibited crease structures indicating plastic deformation. Endogenous growth was prevalent at periods of low discharge rate, notably in the later stages of the eruption, and the style of endogenous growth was compared to the formation of basaltic lava pillows on a much larger scale (Nakada et al. 1995). The overall discharge rate followed a gradually declining trend with time, overprinting a pattern dominated by two distinct pulses (Nakada & Motomura 1999). The style of growth, switching between exogenous and endogenous activity, broadly paralleled this trend. At times of relative quiescence, a viscous plug would develop in the upper conduit, probably in response to shallow degassing and microlite crystallization. Groundmass crystallization of the magma below this plug raised the overpressure beneath the plug to a critical threshold and extruded the viscous plug, marking the onset of a fresh pulse in activity (Nakada & Motomura 1999). A similar mechanism was invoked to explain the extrusion of a large spine onto the endogenous dome summit of Mount Unzen in Decem- ber 1994, an event that signalled the end of the eruption. Thus degassing, groundmass crystallization and consequent rheological stiffening were also key influences at Mount Unzen, as proposed here for the Soufri6re Hills lava dome.

The short eruption in 1989-1990 of Redoubt Volcano, Alaska, involved the rapid growth and destruction of 13 silicic-andesite domes (Miller 1994). The remote and hazardous location of this volcano, perched precariously within a steep amphitheatre, ham- pered visual observations. However, the blocky nature of the domes, and vigorous eruptive degassing, indicate that similar ascent and emplacement mechanisms to those detailed here may have been operating at Redoubt.

The 1951-1952 dome-forming eruption of Mount Lamington in Papua New Guinea involved the development of structures similar to those documented in the Soufri6re Hills eruption. After an initial paroxysmal explosion, dome growth involved the rapid extrusion of near-solid lava constructing a broad dome (Taylor 1958). In the later eruptive stages, localized extrusions took place in a piece- meal fashion, forming multiple peaks on the dome. This process of directed extrusion, and the late-stage formation of large 'hogs-back' features, bears a distinct resemblance to the growth mechanisms described here.

Observations by Perret (1935) throughout the 1929-1932 eruption of Mont Pel6e (Martinique) also highlights similarities to the formation of the Soufri6re Hills dome. Perret documented the gradual construction of a broad lobe at the head of the Rivi6re S6che through a combination of direct observation and photo- graphy. From close proximity, he was able to describe the growth of individual spines and detail their variable nature, with solid outer faces and viscous interiors. Eruptive activity did not involve any




(a)

(b)

400

350

"=>, 300

250

~" 200

150 "O 100

s0

0

Retarded extrusion of

(C) 17 May 1997 lobe ~11 ~ 25 June collapse + rapid extrusion Ill o, 17 May 1997 lobe

. . . . . . . . J I / Development of o~nll~al~1~;Vle;~P ,%ebn ~ IIi l~ / 7 June 1997 obe

. h A , .,=nof ......... . ~ ~ ~' ~.,L 17 May 1997 I ~ i l .... I V ~ .... il;;r)~ 11111111_. j

1 May May June

Date (1997)

July 31 July

Fig. 35. (a) Development of the dome in the build-up to the 25 June 1997 dome collapse. This schematic view is from high above the eastern dome flanks looking westwards. This period of activity was marked by poor visibility and only limited views of the summit area were possible. Shaded area is the active 17 May 1997 lobe. (b) Following the dome collapse on 25 June 1997, rapid development of the 27 June 1997 lobe (shaded) was apparent infilling the collapse amphitheatre. The steep flanks on the northern side had been reconstructed by 7 July 1997, as seen during a brief window of good visibility on that day. (c) Graph highlighting the anti-correlation between periods of rapid lobe growth with the frequency of hybrid earthquakes for the period May to July 1997. Note that the 25 June 1997 dome collapse occurred during a period of intense hybrid earthquake activity such that individual hybrid events were merging together and could not be recorded separately. For this reason, the collapse event does not appear to occur at a period of peak earthquake activity.

large dome collapses or switches in activity, as have characterized Soufri~re Hills. However, the directed extrusion of crystal-rich andesitic spines and broad lobes on the southern flanks of Mont Pel6e suggest similar controlling mechanisms on dome growth.

Perhaps the closest analogy to Soufri~re Hills is one of the most active lava domes in recent history, namely Mount Merapi on the island of Java, Indonesia. Throughout the twentieth century, intermittent activity involved the gradual effusive growth and partial collapse of the summit lava dome, occasionally causing significant loss of life. A wealth of descriptive data exists for its previous eruptive activity, as summarized in Voight et al. (2000). Merapi is noted for the formation of individual lobes, variable scales of dome collapse, and its ability to switch the focus of activity to a different sector of the edifice. Observations at the summit indicate the presence of large, elliptical-shaped lobes of crystal-rich andesitic

lava nestled within horseshoe-shaped collapse scars. All of these features have been documented during the Soufri6re Hills eruption.

Conclusions

This eruption provided invaluable improvements to our understanding of the processes operating during the ascent and emplacement of crystal-rich intermediate lavas. It has also highlighted the need for further research to elucidate the links between dome growth and associated seismic manifestations (e.g. Hidayat et al.

2000), that may result in even better real-time diagnostic capability. The recognition of directed extrusion along shear zones helps to explain the common occurrence of dome collapses and debris- avalanche-forming sector collapses that are a prominent feature




of andesitic volcanism. Indeed, theories described here have also advanced our understanding of the mechanisms that control dome collapse and the triggering of dome-collapse pyroclastic flows. The most significant observations are summarized below.

D. Swanson and T. Druitt for their constructive reviews. This paper is dedicated to the remarkable spirit and resilience of the Montserratian population and to the memory of the nineteen people who lost their lives on 25 June 1997. R.S.J.S. acknowledges a NERC Professorship and R.B.W. a University of Bristol studentship. Published by permission of the Director of the British Geological Survey.

Controls on ascent and emplacement o f andesitic

lava-degassing versus cooling

The evidence from the Soufribre Hills eruption indicates that the processes operating during shallow magma ascent are so effective that the Soufri6re Hills andesite extrudes as one of the most viscous forms of lava on the Earth's surface. Degassing and subsequent microlite crystallization of the crystal-rich andesitic magma feeding the dome operate such that significant rheologic stiffening of the magma occurs. Emplacement switches between hot, crystalline material with considerable strength (at slow ascent rates) and a viscous Newtonian fluid (at faster ascent rates). Cooling plays only a minor role in the emplacement of crystal-rich andesitic lava.

Morphologic variation during emplacement

The growth of the dome was predominantly governed by extrusions of near-solid spines at low discharge rates and shear lobes at moderate discharge rates. On emplacement, these structures were bounded by ductile shear faults that possibly correspond to the sidewalls of the main conduit. Lava was emplaced at the dome summit in the form of these structures, which sporadically broke up into blocky areas during growth or at the post-emplacement stage. Infrequently, some lateral spreading of more mobile, fresh magma was evident, forming a pancake-like, smooth-surfaced lobe, such as the 25 December 1996 lobe. The higher gas content and lower crystal content of the magma at the head of a fresh pulse could be important factors in explaining more fluid-like behaviour.

Structural controls on emplacement

The observations described support the stick-slip mechanisms of dome growth proposed by Denlinger & Hoblitt (1999), Voight et al. (1999) and Wylie et al. (1999). In essence, the shallow ascent of magma involves it being pushed to the surface in a piston-like manner by magmatic pressure. Surface extrusion of the lava was accommodated along shear faults, either in a vertical manner (as spines) or subhorizontally away from the central conduit (as shear lobes) or as blocky lava often during growth spurts. Mapping of these structures highlighted no clear pattern as to the direction in which emplacement occurred. A simple theory is that the extrusion takes place along the path of least resistance (e.g. emplacement of the 22 October 1997 lobe into the open breach of the crater). The orientation of the curved faults is governed by stress distributions, as proposed by Donnadieu & Merle (1998). A new direction of faulting developed when a pulse of fresh, less viscous magma was impeded by a plug of crystalline, stagnated lava. The imposing size of these structures, the complex manner in which they are emplaced, and their ability to stagnate and switch the focus of extrusion pose problems in forecasting hazards during the active stages of such eruptions. The destructive potential of shear lobes (as demonstrated by growth of the 27 March 1997 lobe and its demolition of the southern dome flanks) deserves consideration in monitoring future dome-forming eruptions.

The authors would like to thank all staff of the Montserrat Volcano Observatory, especially members of Team Volume and Team Seismic throughout the eruption. Special mention goes to B. Darroux of the Montserrat Police Force who is responsible for the impressive photographic collection, particularly in the early eruptive stages. Thanks also to G. Skerritt of the Surveys Department who collected most of the theodolite data listed in Table 3, and to S. Powell at the Imaging Unit, Department of Earth Sciences, University of Bristol. We also acknowledge J. Fink,

References

ALIDIBIROV, M. & DINGWELL, D. B. 2000. Three fragmentation mechanisms for highly viscous magma under rapid decompression. Journal of Volcanology and Geothermal Research, 100, 413-421.

ANDERSON, S. W. & FINK, J. H. 1990. The development and distribution of surface textures at the Mount St. Helens dome. In: FINK, J. H. (ed.) Lava Flows and Domes, IA VCEI Proceedings in Volcanology, 2. Springer-Verlag, New York, 25-46.

ANDERSON, S. W. & F~NK, J. H. 1992. Crease structures indicators of emplacement rates and surface stress regimes of lava flows. Geological Society of America Bulletin, 104, 615-625.

BLUNDY, J. & CASHMAN, K. 2001. Ascent-driven crystallisation of dacite magmas at Mount St. Helens 1980-1986. Contributions to Mineralogy and Petrology, 140, 631-650.

CALDER, E. S., LUCKETT, R., SPARKS, R. S. J. & VOIGHT, B. 2002. Mechan- isms of lava dome instability and generation of rockfalls and pyroclastic flows at Soufri6re Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKE- laar, B. P. (eds) The Eruption of SoufriOre Hills' Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 173-190.

CASHMAN, K. V. 1992. Groundmass crystallization of Mount St. Helens dacite 1980-1986 - a tool for interpreting shallow magmatic processes. Contributions to Mineralogy and Petrology, 109, 431-449.

CASHMAN, K. & BLUNDY, J. 2000. Degassing and crystallization of ascend- ing andesite and dacite. Philosophical Transactions of The Royal Society, 358, 1487-1513.

COLE, P. D., CALDER, E. S., SPARKS, R. S. J. Er AL. 2002. Deposits from dome-collapse and fountain-collapse pyroclastic flows at Soufri~re Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of SoufriOre Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 231-262.

DAAG, A. S., DOLAN, M. T., LAGUERTA, E., MEEKER, G. P., NEWHALL, C. G., PALLISTER, J. S. & SOLIDUM, R. 1996. Growth of a postclimactic lava dome at Pinatubo Volcano, July-October 1992. In: NEWHALL, C. & PUNONGBAYAN, R. (eds) Fire and Mud." Eruptions and Lahars of Mount Pinatubo, Philippines. University of Washington Press, Seattle, 647-664.

DENLINGER, R. P. & HOBLITT, R. P. 1999. Cyclic eruptive behaviour of silicic volcanoes. Geology, 27, 459-462.

DEVlNE, J. D., MURPHY, M. D., RUTHERFORD, M. J. ETAL. 1998. Petrologic evidence for pre-eruptive pressure-temperature conditions, and recent reheating, of andesitic magma erupting at the SoufriSre Hills Volcano, Montserrat, West Indies. Geophysical Research Letters, 25, 3669-3672.

DINGWELL, D. B., ROMANO, C. & HESS, K.-U. 1996. The effect of water on the viscosity of a haplogranitic melt under P-T-X conditions relevant to silicic volcanism. Contributions to Mineralogy and Petrology, 124, 19-28.

DONNADIEU, D. & MERLE, O. 1998. Experiments on the indentation process during cryptodome intrusions: New insights into Mount St. Helens deformation. Geology, 26, 79-82.

DRUITT, T. H., YOUNG, S. R., BAPTIE, B. ET AL. 2002. Episodes of cyclic Vulcanian explosive activity with fountain collapse at Soufribre Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of SoufriOre Hills" Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 281-306.

FINK, J. H. & GRIFnTHS, R. W. 1992. A laboratory analog study of the morphology of lava flows extruded from point and line sources. Journal of Volcanology and Geothermal Research, 54, 19-32.

FINK, J. H. & GRIFFITHS, R. W. 1998. Morphology, eruption rates, and rheology of lava domes: Insights from laboratory models. Journal of Geophysical Research, 103, 527-545.

GARDNER, C. A., CASHMAN, K. V. & NEAL, C. A. 1998. Tephra-fall deposits from the 1992 eruption of Crater Peak, Alaska: implications of clast textures for eruptive processes. Bulletin of Volcanology, 59, 537-555.

GRIFFITHS, R. W. 2000. The dynamics of lava flows. Annual Review of Fluid Mechanics, 32, 477-518.

GRIFFITHS, R. W. & FINK, J. H. 1997. Solidifying Bingham extrusions: a model for the growth of silicic lava domes. Journal of Fluid Mech- anics, 347, 13-36.




HAMMER, J. E., CASHMAN, K. V. & VOIGHT, B. 2000. Magmatic processes revealed by textural and compositional trends in Merapi dome lavas. Journal of Volcanology and Geothermal Research, 100, 165-192.

HARFORD, C. L. & SPARKS, R. S. J. 2001. Recent remobilisation of shallow- level intrusions on Montserrat revealed by hydrogen isotope composition of amphiboles. Earth and Planetary Sciences Letters, 185,285-297.

HARFORD, C. L., PRINGLE, M. S., SPARKS, R. S. J. & YOUNG, S. R. 2002. The volcanic evolution of Montserrat using 4~ geochronology. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of SoufriOre Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 93-113.

HIDAYAT, D., VOIGHT, B., LANGSTON, C., RATDOMOPU RBO, A. & EBELING, C. 2000. Broadband seismic experiment at Merapi Volcano, Java, Indo- nesia: very-long-period pulses embedded in multiphase earthquakes. Journal of Volcanology and Geothermal Research, 100, 215-231.

IVERSON, R. M. 1990. Lava domes modeled as brittle shells that enclose pressurized magma, with application to Mount St Helens. In: FINK, J. H. (ed.) Lava Flows and Domes, IAVCEI Proceedings in Volcanology. Springer-Verlag, New York, Vol. 2, 47-69.

LEJEUNE, A.-M. & RICHET, P. 1995. Rheology of crystal-bearing silicate melts: An experimental study at high viscosities. Journal of Geophysical Research, 100, 4215-4229.

LIPMAN, P. W., BANKS, N. G. & RHODES, J. M. 1985. Degassing-induced crystallization of basaltic magma and effects on lava rheology. Nature, 317, 604-607.

LOUGHLIN, S. C., CALDER, E. S., CLARKE, A. B. ET AL. 2002. Pyroclastic flows and surges generated by the 25 June 1997 dome collapse, SoufriGre Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufrikre Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 191-209.

MELNIK, O. & SPARKS, R. S. J. 1999. Nonlinear dynamics of lava dome extrusion. Nature, 402, 37-41.

MELNIK, O. & SPARKS, R. S. J. 2002. Dynamics of magma ascent and lava extrusion at SoufriGre Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufridre Hills Volcano. Mont- serrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 153-171.

MILLER, A. D., STEWART, R. C., WHITE, R. A. ET AL. 1998. Seismicity associated with dome growth and collapse at the Soufri&e Hills Volcano, Montserrat. Geophysical Research Letters, 25, 3401-3404.

MILLER, T. P. 1994. Dome growth and destruction during the 1989-1990 eruption of Redoubt Volcano. Journal of Volcanology and Geothermal Research, 62, 197-212.

MURPHY, M. D., SPARKS, R. S. J., BARCLAY, J., CARROLL, M. R. & BREWER, T. S. 2000. Remobilization of andesite magma by intrusion of mafic magma at the SoufriOre Hills Volcano, Montserrat, West Indies. Journal of Petrology, 41, 21-42.

NAKADA, S. & MOTOMURA, Y. 1999. Petrology of the 1991-1995 eruption at Unzen: effusion pulsation and groundmass crystallization. Journal of Volcanology and Geothermal Research, 89, 173-196.

NAKADA, N., MIYAKE, Y., SATO, H., OSHIMA, 0. & FUJINAWA, A. 1995. Endogenous growth of the dacite dome at Unzen Volcano, Japan 1993-1994. Geology, 23, 157-160.

NAKADA, S., SHIMIZU, H. & OHTA, K. 1999. Overview of the 1990-1995 eruption at Unzen Volcano. Journal o[" Volcanology and Geothermal Research, 89, 1-22.

NEWHALL, C. G. & MELSON, W. G. 1983. Explosive activity associated with the growth of volcanic domes. Journal o[ Volcanology and Geothermal Research, 17, 111 131.

NORTON, G. E., WATTS, R. B., VOIGHT, B. ETAL. 2002. Pyroclastic flow and explosive activity at SoufriGre Hills Volcano, Montserrat, during a period of virtually no magma extrusion (March 1998 to November 1999). In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufridre Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 467-481.

PERRET, F. A. 1935. The Eruption of Mt. PelOe 1929-1932. Carnegie Institution of Washington, Baltimore.

PINKERTON, H. & STEVENSON, R. J. 1992. Methods of determining the rheological properties of magmas at sub-liquidus temperatures. Journal of Volcanology and Geothermal Research, 53, 47-66.

ROBERTSON, R. E. A., COLE, P. D., SPARKS, R. S. J. ET AL. 1998. The explosive eruption of Soufri&e Hills Volcano, Montserrat, West Indies, September 17, 1996, Geophysical Research Letters, 25, 3249-3432.

ROBERTSON, R. E. A., ASPINALL, W. P., HERD, R. A., NORTON, G. E., SPARKS, R. S. J. & YOUNG, S. R. 2000. The 1995-1998 eruption of the SoufriGre Hills Volcano, Montserrat, WI. Philosophical Transactions of the Royal Society, 358, 1619-1637.

ROOBOL, M. J. & SMITH, A. L. 1998. Pyroclastic stratigraphy of the Sou- friGre Hills Volcano, Montserrat: implications for the present eruption. Geophysical Research Letters, 25, 3392-3396.

SPARKS, R. S. J. 1997. Causes and consequences of pressurisation in lava dome eruptions. Earth and Planetary Science Letters, 150, 177-189.

SPARKS, R. S. J. & PINKERTON, H. 1978. Effects of degassing on rheology of basaltic lava. Nature, 276, 385-386.

SPARKS, R. S. J., YOUNG, S. R., BARCLAY, J. ET AL. 1998. Magma production and growth of the lava dome of the Soufri6re Hills Vol- cano, Montserrat, West Indies: November 1995 to December 1997. Geophysical Research Letters, 25, 3421-3424.

SPARKS, R. S. J., MURPHY, M. D., LEJEUNE, A. M., WATTS, R. B., BAR- CLAY, J. & YOUNG, S. R. 2000. Control on the emplacement of the andesite lava dome of the Soufri&e Hills volcano, Montserrat by degassing-induced crystallization. Terra Nova, 12, 14-20.

SPARKS, R. S. J., BARCLAY, J., CALDER, E. S. ET AL. 2002. Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at SoufriGre Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of SoufriOre Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 409-434.

STASIUK, M. V. & JAUPART, C. 1997. Lava flow shapes and dimensions as reflections of magma system conditions. Journal of Volcanology and Geothermal Research, 78, 31-50.

SWANSON, R. A. & HOLCOMB, R. T. 1990. Regularities in growth of the Mount St Helens dacite dome 1980-1986. In: FINK, J. H. (ed.) Lava Flows and Domes. IA VCEI Proceedings in Volcanology. Springer-Verlag, New York, Vol. 2, 3-24.

SWANSON, D. A., DZURISIN, D., HOLCOMB, R. T. ET AL. 1987. Growth of the lava dome at Mount St Helens, Washington, (USA) 1981-1983. In: FINK, J. H. (ed.) The Emplacement of Silicic Domes and Lava Flows. Geological Society of America, Boulder, Special Paper, 212, 1-16.

TAYLOR, G. A. M. 1958. The 1951 eruption of Mount Lamington, Papua. Commonwealth of Australia, Department of National Development, Bureau of Mineral Resources, Geology and Geophysics, Canberra.

VOIGHT, B., SPARKS, R. S. J., MILLER, A. D. ET AL. 1999. Magma flow instability and cyclic activity at Soufri+re Hills Volcano, Montserrat, British West Indies. Science, 283, 1138-1142.

VOIGHT, B., CONSTANTINE, E. K., SISWOWIDJOYO, S. & TORLEY, R. 2000. Historical eruptions of Merapi Volcano, Central Java, Indonesia, 1768-1998. Journal of Volcanology and Geothermal Research, 100, 69-138.

VOIGHT, B., KOMOROWSKI. J.-C., NORTON, G. E. ETAL. 2002. The 26 Decem- ber (Boxing Day) 1997 sector collapse and debris avalanche at SoufriGre Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufrigre Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21,363-407.

WOLF, K. J. & EICHELBERGER, J. C. 1997. Syneruptive mixing, degassing, and crystallization at Redoubt Volcano, eruption of December 1989 to May 1990. Journal of Volcanology and Geothermal Research, 75, 19-37.

WYLIE, J. J., VOIGHT, B. & WHITEHEAD, J. A. 1999. Instability of magma flow from volatile-dependent viscosity. Science, 285, 1883-1885.

YOUNG, S. R., SPARKS, R. S. J., ASP1NALL, W. P., LLYNCH, L. L., MILLER, A. D., ROBERTSON, R. E. A. & SHEPHERD, J. B. 1998. Overview of the eruption of Soufri&e Hills Volcano, Montserrat, July 18 1995, to December 1997. Geophysical Research Letters, 25, 3389-3392.

YOUNG, S. R., VOIGHT, B., BARCLAY, J. ETAL. 2002. Hazard implications of small-scale edifice instability and sector collapse: a case history from SoufriGre Hills Volcano, Montserrat. In: DRU~TT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufridre Hills' Voh'ano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 349-361.



growth patterns and emplacement of the andesitic lava dome...

Documents